[
https://issues.apache.org/jira/browse/AMBARI-21902?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ayub Pathan updated AMBARI-21902:
---------------------------------
Description:
Atlas is connecting to a host where zookeeper is not installed, thinking that
there is zookeeper instance running on that host. Ambari is setting this
property "atlas.graph.storage.hostname" to hbase hostname, instead it should
set to zookeeper quorum. Since Hbase is managed by zookeeper, this setting
should be fine.
*Please note, this issue will not be seen if Hbase master and zookeeper are
hosted on the same node.*
To explain in detail:
* There is a configuration property(atlas.graph.storage.hostname) in Atlas
which defines where Hbase storage is installed.
* This property is managed by Ambari and it is set to the hostname where HBase
is installed.
* When Atlas starts, it initializes the connection with HBase using
HBaseStoreManger class, which overrides “hbase.zookeeper.quorum” property value
with “atlas.graph.storage.hostname” property value.
* If there is zookeeper node running on the same host as HBase host, then the
HBase initialization will succeed, everything is fine.
* If there is no zookeeper node is running on the same host as HBase, then the
above zookeeper connection issue can be observed, which will eventually result
in Atlas failure.
To fix this, “atlas.graph.storage.hostname” config property in Atlas(managed by
Ambari) can be set to the cluster zookeeper quorum value. Since HBase is
managed by Zookeeper, this setting should be fine.
Fix can be done here:
https://github.com/apache/ambari/blob/branch-2.6/ambari-server/src/main/java/org/apache/ambari/server/controller/internal/BlueprintConfigurationProcessor.java#L2734
{noformat}
2017-09-06 23:17:20,518 INFO - [main:] ~ Instantiated HBase compatibility
layer supporting runtime HBase version 1.1.2.2.6.3.0-61:
com.thinkaurelius.titan.diskstorage.hbase.HBaseCompat1_1 (HBaseCompatLoader:69)
2017-09-06 23:17:20,560 INFO - [main:] ~ Copied host list from
root.storage.hostname to hbase.zookeeper.quorum:
ctr-e134-1499953498516-130328-01-000003.hwx.site (HBaseStoreManager:320)
2017-09-06 23:17:20,666 INFO - [main:] ~ Process
identifier=hconnection-0x80b122b connecting to ZooKeeper
ensemble=ctr-e134-1499953498516-130328-01-000003.hwx.site:2181
(RecoverableZooKeeper:120)
{noformat}
was:
Atlas is connecting to a host where zookeeper is not installed, thinking that
there is zookeeper instance running on that host. Ambari is setting this
property "atlas.graph.storage.hostname" to hbase hostname, instead it should
set to zookeeper quorum. Since Hbase is managed by zookeeper, this setting
should be fine.
*Please note, this issue will not be seen if Hbase master and zookeeper are
hosted on the same node.*
To explain in detail:
* There is a configuration property(atlas.graph.storage.hostname) in Atlas
which defines where Hbase storage is installed.
* This property is managed by Ambari and it is set to the hostname where HBase
is installed.
* When Atlas starts, it initializes the connection with HBase using
HBaseStoreManger class, which overrides “hbase.zookeeper.quorum” property value
with “atlas.graph.storage.hostname” property value.
* If there is zookeeper node running on the same host as HBase host, then the
HBase initialization will succeed, everything is fine.
* If there is no zookeeper node is running on the same host as HBase, then the
above zookeeper connection issue can be observed, which will eventually result
in Atlas failure.
To fix this, “atlas.graph.storage.hostname” config property in Atlas(managed by
Ambari) should always be set to the cluster zookeeper quorum value. Since HBase
is managed by Zookeeper, this setting should be fine.
Fix can be done here:
https://github.com/apache/ambari/blob/branch-2.6/ambari-server/src/main/java/org/apache/ambari/server/controller/internal/BlueprintConfigurationProcessor.java#L2734
{noformat}
2017-09-06 23:17:20,518 INFO - [main:] ~ Instantiated HBase compatibility
layer supporting runtime HBase version 1.1.2.2.6.3.0-61:
com.thinkaurelius.titan.diskstorage.hbase.HBaseCompat1_1 (HBaseCompatLoader:69)
2017-09-06 23:17:20,560 INFO - [main:] ~ Copied host list from
root.storage.hostname to hbase.zookeeper.quorum:
ctr-e134-1499953498516-130328-01-000003.hwx.site (HBaseStoreManager:320)
2017-09-06 23:17:20,666 INFO - [main:] ~ Process
identifier=hconnection-0x80b122b connecting to ZooKeeper
ensemble=ctr-e134-1499953498516-130328-01-000003.hwx.site:2181
(RecoverableZooKeeper:120)
{noformat}
> Ambari is setting "atlas.graph.storage.hostname" to hbase hostname, instead
> it should set the zookeeper quorum.
> ---------------------------------------------------------------------------------------------------------------
>
> Key: AMBARI-21902
> URL: https://issues.apache.org/jira/browse/AMBARI-21902
> Project: Ambari
> Issue Type: Bug
> Components: ambari-server
> Affects Versions: 2.5.0, 2.6.0
> Reporter: Ayub Pathan
> Assignee: Ayub Pathan
> Priority: Critical
> Fix For: 2.6.0
>
> Attachments: AMBARI-21902.patch
>
>
> Atlas is connecting to a host where zookeeper is not installed, thinking that
> there is zookeeper instance running on that host. Ambari is setting this
> property "atlas.graph.storage.hostname" to hbase hostname, instead it should
> set to zookeeper quorum. Since Hbase is managed by zookeeper, this setting
> should be fine.
> *Please note, this issue will not be seen if Hbase master and zookeeper are
> hosted on the same node.*
> To explain in detail:
> * There is a configuration property(atlas.graph.storage.hostname) in Atlas
> which defines where Hbase storage is installed.
> * This property is managed by Ambari and it is set to the hostname where
> HBase is installed.
> * When Atlas starts, it initializes the connection with HBase using
> HBaseStoreManger class, which overrides “hbase.zookeeper.quorum” property
> value with “atlas.graph.storage.hostname” property value.
> * If there is zookeeper node running on the same host as HBase host, then the
> HBase initialization will succeed, everything is fine.
> * If there is no zookeeper node is running on the same host as HBase, then
> the above zookeeper connection issue can be observed, which will eventually
> result in Atlas failure.
> To fix this, “atlas.graph.storage.hostname” config property in Atlas(managed
> by Ambari) can be set to the cluster zookeeper quorum value. Since HBase is
> managed by Zookeeper, this setting should be fine.
>
> Fix can be done here:
> https://github.com/apache/ambari/blob/branch-2.6/ambari-server/src/main/java/org/apache/ambari/server/controller/internal/BlueprintConfigurationProcessor.java#L2734
> {noformat}
> 2017-09-06 23:17:20,518 INFO - [main:] ~ Instantiated HBase compatibility
> layer supporting runtime HBase version 1.1.2.2.6.3.0-61:
> com.thinkaurelius.titan.diskstorage.hbase.HBaseCompat1_1
> (HBaseCompatLoader:69)
> 2017-09-06 23:17:20,560 INFO - [main:] ~ Copied host list from
> root.storage.hostname to hbase.zookeeper.quorum:
> ctr-e134-1499953498516-130328-01-000003.hwx.site (HBaseStoreManager:320)
> 2017-09-06 23:17:20,666 INFO - [main:] ~ Process
> identifier=hconnection-0x80b122b connecting to ZooKeeper
> ensemble=ctr-e134-1499953498516-130328-01-000003.hwx.site:2181
> (RecoverableZooKeeper:120)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
