[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17119153#comment-17119153 ] Bharath Vissapragada commented on HBASE-18095: -- > We have dynamic reloading of configuration implemented in some areas. The > ability to reload the configured master addresses seems like a good place to > make use of that capability. Yes, I saw the whole ConfigurationObserver thing. For server roles, looks like it is triggered by a an updateConfiguration() RPC but in this case we need a periodic refresher that that reloads configuration on the client. > For the interface itself, should it return a Set of java.net.URL or > java.net.SocketAddress instead? Right, or a HostAndPort... I can spin up a patch unless there are any objections with this approach. > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0-alpha-1, 2.3.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17117912#comment-17117912 ] Nick Dimiduk commented on HBASE-18095: -- We have dynamic reloading of configuration implemented in some areas. The ability to reload the configured master addresses seems like a good place to make use of that capability. As for extending the interface, that would be fine, if a little heavy-handed. I prefer it be a separate interface, as you suggest, so that it can be swapped out without having to re-implement the entire registry. For the interface itself, should it return a {{Set}} of {{java.net.URL}} or {{java.net.SocketAddress}} instead? > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0-alpha-1, 2.3.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17113440#comment-17113440 ] Bharath Vissapragada commented on HBASE-18095: -- > Maybe we could add an option called fallbackToZk? Implementation wise, I think it's pretty simple to do that. We can have a "compound" registry implementation that delegates to primary and secondary registries but yes if all the masters are down and ZK servers are available, the cluster is already in a bad shape. I think there is very little a client could with the connection. That said, after following the discussion here and in HBASE-11288, I think what the current registry implementations lack (out of the box) is some sort of dynamic resolver interface and re-configuration on the fly. Currently I think everyone is exploiting their infrastructure specific solutions (DNS resolution, lvs etc) to trick the rpcs to be load balanced. Was wondering if we can abstract some of that out and make it pluggable in the client code. For example, a simple interface like {code:java} interface RegistryEndPointResolver { Set getRegistryEndPoints(); } {code} The default implementation would be based on the hbase-site.xml config and polled every few seconds. But every one can plug their own implementation into it based on say, DNS, existing service discovery solutions etc. Thoughts? > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0-alpha-1, 2.3.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17113290#comment-17113290 ] Andrew Kyle Purtell commented on HBASE-18095: - One objective of this work is to remove dependencies on ZK in the client for the normal client cases. That objective is defeated if there is a fallback. It would have to be optional. Now we have an optional fallback for an optional choice to remove dependencies on ZK that reintroduces a dependency on ZK? I don’t believe it provides much value anyway. It’s the same situation as if all your ZK quorum goes down. Someone or something (like an autoscaler or process supervisor) will have to start new instances in either case. > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0-alpha-1, 2.3.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17113072#comment-17113072 ] Duo Zhang commented on HBASE-18095: --- IIRC, one of the problem for MasterRegistry is that, if all the configured masters are gone, then we need to change the configuration to let client know the new masters. Maybe we could add an option called fallbackToZk? If enabled, when there are no masters available, we fallback to use the zookeeper to fetch all the active and backup masters and then update our configuration. Thoughts? Thanks. > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0-alpha-1, 2.3.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17046380#comment-17046380 ] HBase QA commented on HBASE-18095: -- | (x) *{color:red}-1 overall{color}* | \\ \\ || Vote || Subsystem || Runtime || Comment || | {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 1m 28s{color} | {color:blue} Docker mode activated. {color} | || || || || {color:brown} Prechecks {color} || | {color:green}+1{color} | {color:green} dupname {color} | {color:green} 0m 2s{color} | {color:green} No case conflicting files found. {color} | | {color:blue}0{color} | {color:blue} prototool {color} | {color:blue} 0m 0s{color} | {color:blue} prototool was not available. {color} | | {color:green}+1{color} | {color:green} hbaseanti {color} | {color:green} 0m 0s{color} | {color:green} Patch does not have any anti-patterns. {color} | | {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m 0s{color} | {color:green} The patch does not contain any @author tags. {color} | | {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m 0s{color} | {color:green} The patch appears to include 97 new or modified test files. {color} | || || || || {color:brown} branch-2 Compile Tests {color} || | {color:blue}0{color} | {color:blue} mvndep {color} | {color:blue} 0m 14s{color} | {color:blue} Maven dependency ordering for branch {color} | | {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 5m 51s{color} | {color:green} branch-2 passed {color} | | {color:green}+1{color} | {color:green} compile {color} | {color:green} 3m 18s{color} | {color:green} branch-2 passed {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 3m 7s{color} | {color:green} branch-2 passed {color} | | {color:green}+1{color} | {color:green} shadedjars {color} | {color:green} 5m 10s{color} | {color:green} branch has no errors when building our shaded downstream artifacts. {color} | | {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 2m 21s{color} | {color:green} branch-2 passed {color} | | {color:blue}0{color} | {color:blue} spotbugs {color} | {color:blue} 0m 36s{color} | {color:blue} Used deprecated FindBugs config; considering switching to SpotBugs. {color} | | {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 11m 10s{color} | {color:green} branch-2 passed {color} | || || || || {color:brown} Patch Compile Tests {color} || | {color:blue}0{color} | {color:blue} mvndep {color} | {color:blue} 0m 14s{color} | {color:blue} Maven dependency ordering for patch {color} | | {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 5m 26s{color} | {color:green} the patch passed {color} | | {color:green}+1{color} | {color:green} compile {color} | {color:green} 3m 36s{color} | {color:green} the patch passed {color} | | {color:green}+1{color} | {color:green} cc {color} | {color:green} 3m 36s{color} | {color:green} the patch passed {color} | | {color:green}+1{color} | {color:green} javac {color} | {color:green} 3m 36s{color} | {color:green} the patch passed {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m 10s{color} | {color:green} The patch passed checkstyle in hbase-protocol-shaded {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m 27s{color} | {color:green} hbase-common: The patch generated 0 new + 16 unchanged - 2 fixed = 16 total (was 18) {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m 42s{color} | {color:green} hbase-client: The patch generated 0 new + 262 unchanged - 24 fixed = 262 total (was 286) {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m 13s{color} | {color:green} The patch passed checkstyle in hbase-zookeeper {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 1m 24s{color} | {color:green} hbase-server: The patch generated 0 new + 640 unchanged - 68 fixed = 640 total (was 708) {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m 13s{color} | {color:green} The patch passed checkstyle in hbase-endpoint {color} | | {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m 0s{color} | {color:green} The patch has no whitespace issues. {color} | | {color:green}+1{color} | {color:green} shadedjars {color} | {color:green} 4m 42s{color} | {color:green} patch has no errors when building our shaded downstream artifacts. {color} | | {color:green}+1{color} | {color:green} hadoopcheck {color} | {color:green} 17m 46s{color} | {color:green} Patch does not cause any errors with Hadoop 2.8.5 2.9.2 or 3.1.2. {color} | | {color:green}+1{color} | {color:green} hbaseprotoc {color} | {color:green} 3m 2s{color} |
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17046132#comment-17046132 ] HBase QA commented on HBASE-18095: -- | (x) *{color:red}-1 overall{color}* | \\ \\ || Vote || Subsystem || Runtime || Comment || | {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 1m 24s{color} | {color:blue} Docker mode activated. {color} | || || || || {color:brown} Prechecks {color} || | {color:green}+1{color} | {color:green} dupname {color} | {color:green} 0m 2s{color} | {color:green} No case conflicting files found. {color} | | {color:blue}0{color} | {color:blue} prototool {color} | {color:blue} 0m 0s{color} | {color:blue} prototool was not available. {color} | | {color:green}+1{color} | {color:green} hbaseanti {color} | {color:green} 0m 0s{color} | {color:green} Patch does not have any anti-patterns. {color} | | {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m 0s{color} | {color:green} The patch does not contain any @author tags. {color} | | {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m 0s{color} | {color:green} The patch appears to include 97 new or modified test files. {color} | || || || || {color:brown} branch-2 Compile Tests {color} || | {color:blue}0{color} | {color:blue} mvndep {color} | {color:blue} 0m 14s{color} | {color:blue} Maven dependency ordering for branch {color} | | {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 6m 11s{color} | {color:green} branch-2 passed {color} | | {color:green}+1{color} | {color:green} compile {color} | {color:green} 3m 30s{color} | {color:green} branch-2 passed {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 3m 5s{color} | {color:green} branch-2 passed {color} | | {color:green}+1{color} | {color:green} shadedjars {color} | {color:green} 4m 52s{color} | {color:green} branch has no errors when building our shaded downstream artifacts. {color} | | {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 2m 7s{color} | {color:green} branch-2 passed {color} | | {color:blue}0{color} | {color:blue} spotbugs {color} | {color:blue} 0m 37s{color} | {color:blue} Used deprecated FindBugs config; considering switching to SpotBugs. {color} | | {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 10m 29s{color} | {color:green} branch-2 passed {color} | || || || || {color:brown} Patch Compile Tests {color} || | {color:blue}0{color} | {color:blue} mvndep {color} | {color:blue} 0m 14s{color} | {color:blue} Maven dependency ordering for patch {color} | | {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 5m 36s{color} | {color:green} the patch passed {color} | | {color:green}+1{color} | {color:green} compile {color} | {color:green} 3m 27s{color} | {color:green} the patch passed {color} | | {color:green}+1{color} | {color:green} cc {color} | {color:green} 3m 27s{color} | {color:green} the patch passed {color} | | {color:green}+1{color} | {color:green} javac {color} | {color:green} 3m 27s{color} | {color:green} the patch passed {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m 10s{color} | {color:green} The patch passed checkstyle in hbase-protocol-shaded {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m 25s{color} | {color:green} hbase-common: The patch generated 0 new + 16 unchanged - 2 fixed = 16 total (was 18) {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m 39s{color} | {color:green} hbase-client: The patch generated 0 new + 262 unchanged - 24 fixed = 262 total (was 286) {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m 12s{color} | {color:green} The patch passed checkstyle in hbase-zookeeper {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 1m 24s{color} | {color:green} hbase-server: The patch generated 0 new + 640 unchanged - 68 fixed = 640 total (was 708) {color} | | {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m 13s{color} | {color:green} The patch passed checkstyle in hbase-endpoint {color} | | {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m 0s{color} | {color:green} The patch has no whitespace issues. {color} | | {color:green}+1{color} | {color:green} shadedjars {color} | {color:green} 4m 45s{color} | {color:green} patch has no errors when building our shaded downstream artifacts. {color} | | {color:green}+1{color} | {color:green} hadoopcheck {color} | {color:green} 18m 20s{color} | {color:green} Patch does not cause any errors with Hadoop 2.8.5 2.9.2 or 3.1.2. {color} | | {color:green}+1{color} | {color:green} hbaseprotoc {color} | {color:green} 4m 20s{color} |
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17041263#comment-17041263 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #84 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/84/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/84//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/84//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/84//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17040903#comment-17040903 ] Hudson commented on HBASE-18095: Results for branch master [build #1637 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/master/1637/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/master/1637//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/master/1637//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/master/1637//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17040665#comment-17040665 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #83 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/83/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/83//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/83//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/83//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17040612#comment-17040612 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #82 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/82/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/82//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/82//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/82//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.7.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17040376#comment-17040376 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #81 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/81/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/81//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/81//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/81//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.7.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17039406#comment-17039406 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #80 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/80/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/80//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/80//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/80//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.7.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17038609#comment-17038609 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #79 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/79/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/79//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/79//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/79//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.7.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17037952#comment-17037952 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #78 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/78/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/78//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/78//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/78//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.7.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17037582#comment-17037582 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #77 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/77/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/77//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/77//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/77//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.7.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17037405#comment-17037405 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #76 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/76/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/76//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/76//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/76//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.7.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17037204#comment-17037204 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #75 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/75/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/75//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/75//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/75//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.7.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17036465#comment-17036465 ] Andrew Kyle Purtell commented on HBASE-18095: - We are releasing 1.6.0 now because of HBASE-23825. Moving to 1.7.0 > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.7.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17036443#comment-17036443 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #74 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/74/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/74//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/74//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/74//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17035660#comment-17035660 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #73 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/73/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/73//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/73//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/73//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17034780#comment-17034780 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #72 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/72/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/72//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/72//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/72//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17034178#comment-17034178 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #71 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/71/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/71//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/71//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/71//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17034049#comment-17034049 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #70 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/70/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/70//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/70//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/70//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17033994#comment-17033994 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #69 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/69/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/69//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/69//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/69//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/69//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17033102#comment-17033102 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #67 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/67/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/67//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/67//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/67//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/67//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17032790#comment-17032790 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #66 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/66/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/66//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/66//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/66//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/66//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17032069#comment-17032069 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #65 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/65/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/65//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/65//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/65//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/65//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17031199#comment-17031199 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #64 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/64/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/64//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/64//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/64//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/64//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17030493#comment-17030493 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #63 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/63/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/63//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/63//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/63//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/63//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17030251#comment-17030251 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #62 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/62/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/62//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/62//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/62//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/62//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17029529#comment-17029529 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #61 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/61/]: (x) *{color:red}-1 overall{color}* details (if available): (x) {color:red}-1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/61//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/61//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/61//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/61//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17029176#comment-17029176 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #60 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/60/]: (x) *{color:red}-1 overall{color}* details (if available): (x) {color:red}-1 general checks{color} -- Something went wrong running this stage, please [check relevant console output|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/60//console]. (x) {color:red}-1 jdk8 hadoop2 checks{color} -- Something went wrong running this stage, please [check relevant console output|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/60//console]. (x) {color:red}-1 jdk8 hadoop3 checks{color} -- Something went wrong running this stage, please [check relevant console output|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/60//console]. (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/60//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17028635#comment-17028635 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #59 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/59/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/59//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/59//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/59//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/59//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17028250#comment-17028250 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #58 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/58/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/58//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/58//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/58//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/58//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17027941#comment-17027941 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #57 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/57/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/57//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/57//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/57//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/57//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17027134#comment-17027134 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #56 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/56/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/56//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/56//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/56//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/56//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17026398#comment-17026398 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #55 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/55/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/55//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/55//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/55//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/55//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17025621#comment-17025621 ] Bharath Vissapragada commented on HBASE-18095: -- Thanks [~stack]. > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17025599#comment-17025599 ] Michael Stack commented on HBASE-18095: --- [~bharathv] Pushed again. I think you need the recent errorprone fixes HBASE-23753 on your branch too. > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17025538#comment-17025538 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #54 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/54/]: (x) *{color:red}-1 overall{color}* details (if available): (x) {color:red}-1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/54//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/54//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/54//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/54//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17025515#comment-17025515 ] Michael Stack commented on HBASE-18095: --- Rebased [~bharathv] > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17025486#comment-17025486 ] Bharath Vissapragada commented on HBASE-18095: -- [~stack] Sure. I'll put up a discuss thread on dev once the branch is stable. I'm working on fixing on the last set of test failures exposed in nightly (via HBASE-23752). > Does the branch need updating (seems old when I compare to master) Yes please. Can you please rebase it when you get a chance? > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17025478#comment-17025478 ] Michael Stack commented on HBASE-18095: --- Does the branch need updating (seems old when I compare to master) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17025471#comment-17025471 ] Michael Stack commented on HBASE-18095: --- I'd be in favor of master merge after you think it stable out on the branch. Merge discussion should be on dev list FYI. Let me see if can do a quick review of the branch... Hopefully will help. > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17024908#comment-17024908 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #53 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/53/]: (x) *{color:red}-1 overall{color}* details (if available): (x) {color:red}-1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/53//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/53//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/53//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (x) {color:red}-1 client integration test{color} --Failed when running client tests on top of Hadoop 2. [see log for details|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/53//artifact/output-integration/hadoop-2.log]. (note that this means we didn't run on Hadoop 3) > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Fix For: 3.0.0, 2.3.0, 1.6.0 > > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17024678#comment-17024678 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #52 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/52/]: (x) *{color:red}-1 overall{color}* details (if available): (x) {color:red}-1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/52//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/52//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/52//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17024646#comment-17024646 ] Andrew Kyle Purtell commented on HBASE-18095: - Sounds like a plan :-) Since those PRs were approved I merged them. > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17024640#comment-17024640 ] Bharath Vissapragada commented on HBASE-18095: -- For those following this work, I'd like to update the current status. Most critical patches are already in (see the subtasks). There are just two pending PRs [1, 2] that are already approved and should be committed soon. The nightly test runs on the feature branch [3] are relatively stable (except the flaky test of TestFromClientSide which is fixed by [2] and the remaining tests failures are not specific to this branch). Once the remaining patches are committed (that switches the default registry), I can keep an eye on the nightly job for a few days to make sure nothing else is broken. At this point there are no bugs/feature gaps/performance concerns that I'm aware of (please correct me if I'm wrong). Now that we are pretty close to getting all the needed patches in, I'd like to kickstart the discussion of merging the feature branch into master and see what other folks think. It is a pain to maintain both the branches since we need frequent rebases. Given the current status of the work and the nightly test runs and the test coverage, I'm fairly confident that this work can land in the master branch. There might be some very little initial friction (for developers writing unit tests) since the code changes are invasive and touch many commonly used codepaths (like connection setup, RPC framework etc) and also rewrites a bunch of tests and common test utilities. But I think those issues can be treated on a case-by-case basis. Worst case, we have a kill-switch for the entire feature (one line change), if we really need it. If everyone agrees to merge the feature branch, the follow up work would be to - Document the feature (client and developer facing) and expectations - HBASE-23331 - Work on the back ports for branch-2 and branch-1 (targeting the upcoming 2.3.0) - Doing some perf runs to see how it compares with the baseline (ZK based registry) and see how the nightly runs behave - Fixing test coverage gaps if any. [~apurtell] [~ndimiduk] [~stack] thoughts? — References [1] [https://github.com/apache/hbase/pull/1039] [2] [https://github.com/apache/hbase/pull/1091] [3] [https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/] > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17023931#comment-17023931 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #51 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/51/]: (x) *{color:red}-1 overall{color}* details (if available): (x) {color:red}-1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/51//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/51//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/51//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17023652#comment-17023652 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #50 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/50/]: (x) *{color:red}-1 overall{color}* details (if available): (x) {color:red}-1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/50//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/50//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/50//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17023277#comment-17023277 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #49 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/49/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/49//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/49//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/49//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17022504#comment-17022504 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #48 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/48/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/48//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/48//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/48//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17021525#comment-17021525 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #47 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/47/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/47//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/47//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/47//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17020601#comment-17020601 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #46 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/46/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/46//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/46//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/46//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17019749#comment-17019749 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #45 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/45/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/45//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/45//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/45//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17019043#comment-17019043 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #44 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/44/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/44//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/44//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/44//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17018711#comment-17018711 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #43 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/43/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/43//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/43//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/43//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17018333#comment-17018333 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #42 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/42/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/42//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/42//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/42//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17017516#comment-17017516 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #41 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/41/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/41//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/41//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/41//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17016343#comment-17016343 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #40 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/40/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/40//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/40//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/40//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17015407#comment-17015407 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #39 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/39/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/39//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/39//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/39//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17014698#comment-17014698 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #38 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/38/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/38//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/38//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/38//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17013909#comment-17013909 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #37 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/37/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/37//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/37//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/37//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17013579#comment-17013579 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #36 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/36/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/36//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/36//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/36//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17013250#comment-17013250 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #35 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/35/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/35//General_Nightly_Build_Report/] (/) {color:green}+1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/35//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/35//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17012240#comment-17012240 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #34 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/34/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/34//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/34//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/34//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17011030#comment-17011030 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #33 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/33/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/33//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/33//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/33//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17010137#comment-17010137 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #32 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/32/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/32//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/32//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/32//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17009190#comment-17009190 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #31 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/31/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/31//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/31//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/31//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17008430#comment-17008430 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #30 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/30/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/30//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/30//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/30//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17008173#comment-17008173 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #29 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/29/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/29//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/29//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/29//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17007790#comment-17007790 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #28 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/28/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/28//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/28//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/28//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17007088#comment-17007088 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #27 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/27/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/27//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/27//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/27//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17006489#comment-17006489 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #26 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/26/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/26//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/26//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/26//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17006232#comment-17006232 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #25 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/25/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/25//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/25//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/25//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17005067#comment-17005067 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #24 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/24/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/24//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/24//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/24//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17004603#comment-17004603 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #23 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/23/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/23//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/23//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/23//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17004357#comment-17004357 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #22 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/22/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/22//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/22//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/22//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17003792#comment-17003792 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #21 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/21/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/21//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/21//JDK8_Nightly_Build_Report_(Hadoop2)/] (/) {color:green}+1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/21//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17003396#comment-17003396 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #20 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/20/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/20//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/20//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/20//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17003015#comment-17003015 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #19 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/19/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/19//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/19//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/19//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17002526#comment-17002526 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #18 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/18/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/18//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/18//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/18//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17002016#comment-17002016 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #17 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/17/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/17//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/17//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/17//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17001794#comment-17001794 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #16 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/16/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/16//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/16//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/16//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17001301#comment-17001301 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #15 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/15/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/15//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/15//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/15//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17000459#comment-17000459 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #14 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/14/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/14//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/14//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/14//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16999661#comment-16999661 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #13 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/13/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/13//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/13//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/13//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16998689#comment-16998689 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #12 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/12/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/12//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/12//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/12//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16997675#comment-16997675 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #11 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/11/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/11//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/11//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/11//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16996893#comment-16996893 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #10 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/10/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/10//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/10//JDK8_Nightly_Build_Report_(Hadoop2)/] (/) {color:green}+1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/10//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16996581#comment-16996581 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #9 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/9/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/9//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/9//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/9//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16995994#comment-16995994 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #8 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/8/]: (/) *{color:green}+1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/8//General_Nightly_Build_Report/] (/) {color:green}+1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/8//JDK8_Nightly_Build_Report_(Hadoop2)/] (/) {color:green}+1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/8//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16995136#comment-16995136 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #7 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/7/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/7//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/7//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/7//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16993997#comment-16993997 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #6 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/6/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/6//General_Nightly_Build_Report/] (/) {color:green}+1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/6//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/6//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16993948#comment-16993948 ] Michael Stack commented on HBASE-18095: --- [~bharathv] Writeup is excellent. Thanks. Helps. Left a few notes on it if of use. > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16993938#comment-16993938 ] Bharath Vissapragada commented on HBASE-18095: -- [~stack] I wrote a quick google doc to summarize this so that it is easier to collaborate using inline comments. Appreciate any feedback. https://docs.google.com/document/d/1JAJdM7eUxg5b417f0xWS4NztKCx1f2b6wZrudPtiXF4/edit?usp=sharing > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16993865#comment-16993865 ] Michael Stack commented on HBASE-18095: --- [~bharathv] (This is great stuff). Was wondering if there is a one-pager on the overall approach? I'm interested in high-level stuff. For example, how we tell client about list of Masters. Will the client try one Master first or will it query all in parallel or will it be a hedged read. We are adding Masters' inline with read/write now which is a new development. What do we think the loading will be like? Will we need to do work to make the Master fit for new purpose. Thanks. > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16992031#comment-16992031 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #5 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/5/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/5//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/5//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/5//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (HBASE-18095) Provide an option for clients to find the server hosting META that does not involve the ZooKeeper client
[ https://issues.apache.org/jira/browse/HBASE-18095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16991011#comment-16991011 ] Hudson commented on HBASE-18095: Results for branch HBASE-18095/client-locate-meta-no-zookeeper [build #4 on builds.a.o|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/4/]: (x) *{color:red}-1 overall{color}* details (if available): (/) {color:green}+1 general checks{color} -- For more information [see general report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/4//General_Nightly_Build_Report/] (x) {color:red}-1 jdk8 hadoop2 checks{color} -- For more information [see jdk8 (hadoop2) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/4//JDK8_Nightly_Build_Report_(Hadoop2)/] (x) {color:red}-1 jdk8 hadoop3 checks{color} -- For more information [see jdk8 (hadoop3) report|https://builds.apache.org/job/HBase%20Nightly/job/HBASE-18095%252Fclient-locate-meta-no-zookeeper/4//JDK8_Nightly_Build_Report_(Hadoop3)/] (/) {color:green}+1 source release artifact{color} -- See build output for details. (/) {color:green}+1 client integration test{color} > Provide an option for clients to find the server hosting META that does not > involve the ZooKeeper client > > > Key: HBASE-18095 > URL: https://issues.apache.org/jira/browse/HBASE-18095 > Project: HBase > Issue Type: New Feature > Components: Client >Reporter: Andrew Kyle Purtell >Assignee: Bharath Vissapragada >Priority: Major > Attachments: HBASE-18095.master-v1.patch, HBASE-18095.master-v2.patch > > > Clients are required to connect to ZooKeeper to find the location of the > regionserver hosting the meta table region. Site configuration provides the > client a list of ZK quorum peers and the client uses an embedded ZK client to > query meta location. Timeouts and retry behavior of this embedded ZK client > are managed orthogonally to HBase layer settings and in some cases the ZK > cannot manage what in theory the HBase client can, i.e. fail fast upon outage > or network partition. > We should consider new configuration settings that provide a list of > well-known master and backup master locations, and with this information the > client can contact any of the master processes directly. Any master in either > active or passive state will track meta location and respond to requests for > it with its cached last known location. If this location is stale, the client > can ask again with a flag set that requests the master refresh its location > cache and return the up-to-date location. Every client interaction with the > cluster thus uses only HBase RPC as transport, with appropriate settings > applied to the connection. The configuration toggle that enables this > alternative meta location lookup should be false by default. > This removes the requirement that HBase clients embed the ZK client and > contact the ZK service directly at the beginning of the connection lifecycle. > This has several benefits. ZK service need not be exposed to clients, and > their potential abuse, yet no benefit ZK provides the HBase server cluster is > compromised. Normalizing HBase client and ZK client timeout settings and > retry behavior - in some cases, impossible, i.e. for fail-fast - is no longer > necessary. > And, from [~ghelmling]: There is an additional complication here for > token-based authentication. When a delegation token is used for SASL > authentication, the client uses the cluster ID obtained from Zookeeper to > select the token identifier to use. So there would also need to be some > Zookeeper-less, unauthenticated way to obtain the cluster ID as well. -- This message was sent by Atlassian Jira (v8.3.4#803005)