[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15972997#comment-15972997 ] sankalp kohli commented on CASSANDRA-13442: --- Regarding cost, it will help you save roughly 30% if you go from RF=6 to 4(2 in each DC) and more if you do this with RF=10 to 6(3 in each DC). It will be quite complex and we can spec it out to see if it is worth the time. > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15972697#comment-15972697 ] Sylvain Lebresne commented on CASSANDRA-13442: -- bq. Man, this makes me really nervous. bq. Optimizing away redundant queries a la 7168? Sign me up. But I think removing that "redundant" data and making RF not actually mean RF is going too far. I very much second those sentiments. bq. It will be opt in at the keyspace level and won't affect anyone who dont want to use it. Maybe that's meant to be a response to Jonathan's concern I'm quoting (and agreeing with) above. In which case I want to point out that, as also said above, this will imo add quite a bit of complexity to existing code, so I very strongly disagree with the argument that having it opt-in from the user point of view equates it to affecting no-one that don't want to use it. bq. I think it's worth characterizing as much as possible what it's going to cost before dismissing it Sure and I, for one, certainly won't claim having put tons of though into this yet, and I'm happy to see a much more precise analysis of the costs. But I'll admit that my "gut feeling" (my 6+-years-working-on-C*-gut feeling) is that this will get pretty complex especially when you throw up node movements and potential data loss into the mix, and that it's not worth that complexity. > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15972641#comment-15972641 ] T Jake Luciani commented on CASSANDRA-13442: Rather than forcing auto deletion of the data on repair, would you be ok with requiring a explicit cleanup to see the disk savings? That seems like the safest approach. > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[
https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15972633#comment-15972633
]
T Jake Luciani commented on CASSANDRA-13442:
I like the idea of making it part of the replication strategy.
You could have an unrepaired RF and a repaired RF.
In your example it would be: unrepaired: { dc1=3, dc2=3}, repaired: { dc1=2,
dc2=2 }.
> Support a means of strongly consistent highly available replication with
> storage requirements approximating RF=2
>
>
> Key: CASSANDRA-13442
> URL: https://issues.apache.org/jira/browse/CASSANDRA-13442
> Project: Cassandra
> Issue Type: Improvement
> Components: Compaction, Coordination, Distributed Metadata, Local
> Write-Read Paths
>Reporter: Ariel Weisberg
>
> Replication factors like RF=2 can't provide strong consistency and
> availability because if a single node is lost it's impossible to reach a
> quorum of replicas. Stepping up to RF=3 will allow you to lose a node and
> still achieve quorum for reads and writes, but requires committing additional
> storage.
> The requirement of a quorum for writes/reads doesn't seem to be something
> that can be relaxed without additional constraints on queries, but it seems
> like it should be possible to relax the requirement that 3 full copies of the
> entire data set are kept. What is actually required is a covering data set
> for the range and we should be able to achieve a covering data set and high
> availability without having three full copies.
> After a repair we know that some subset of the data set is fully replicated.
> At that point we don't have to read from a quorum of nodes for the repaired
> data. It is sufficient to read from a single node for the repaired data and a
> quorum of nodes for the unrepaired data.
> One way to exploit this would be to have N replicas, say the last N replicas
> (where N varies with RF) in the preference list, delete all repaired data
> after a repair completes. Subsequent quorum reads will be able to retrieve
> the repaired data from any of the two full replicas and the unrepaired data
> from a quorum read of any replica including the "transient" replicas.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15972097#comment-15972097 ] Ariel Weisberg commented on CASSANDRA-13442: bq. if you have three copies and you lose one, no big deal, you still have two to restore from. Just two copies? If anything goes wrong with that other copy while you repair you are SOL I just realized you are getting thrown off by RF=3 N=1. That is just an example. You can run RF=6 N=2 with 3 replicas in each data center and read/write at LOCAL_QUORUM. You still save 1/3 storage. A lot of use cases are massively over replicated when it comes to simply preserving data from hardware failure. Ideally we could use erasure coding, but you still need to make strong consistency happen. I think part of the issue is that there is a conflation of how we determine what the state should be with how we replicate the state. > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15972084#comment-15972084 ] Ariel Weisberg commented on CASSANDRA-13442: bq. 7168 added a new CL as a way to opt-in to this new feature. Once its fully vetted it would be trivial to make it automatically use it when appropriate. I agree it should be opt in. That is why you pick N transient replicas. If N=0 there are only regular replicas. bq. Optimizing away redundant queries a la 7168? Sign me up. But I think removing that "redundant" data and making RF not actually mean RF is going too far. Cassandra will happily start with RF=1 SimpleStrategy. What about providing mechanism not policy? Not all use cases have the same requirements. RF=2 with strong consistency is adequate for some real world use cases where RF=2 without strong consistency would not. bq. >The topology of the cluster would also have a new dimension that the drivers would need to consider. Since for CL.ONE queries you would need to only use one of the replicas with all the data on it. bq. Yes, I think there are going to be multiple places where this gets more complicated than it looks at first. I agree it's not a simple change. I think the question is whether it is feasible and whether the value delivered for storage bound use cases justifies the cost of initial development plus the cost of carrying the functionality into the future. I think it's worth characterizing as much as possible what it's going to cost before dismissing it because being able to run RF=2 with strong consistency is a big deal for storage bound use cases. > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15972055#comment-15972055 ] sankalp kohli commented on CASSANDRA-13442: --- [~tjake] We need to make changes to bootstrap and other operations to know about this. This will also involve changes in the resolver to know which is full replica vs partial replica and make best use of that. > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15972051#comment-15972051 ] sankalp kohli commented on CASSANDRA-13442: --- This will be used with multiple DC having 2 full replicas each. With minimum 2 DCs, you will still have 4 full replicas. Also it can be used with 3 full replicas and 2 partial which can be used for use cases with RF=5 in each DC. It will be opt in at the keyspace level and won't affect anyone who dont want to use it. > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15972046#comment-15972046 ] Jeff Jirsa commented on CASSANDRA-13442: Why does it need to be automatic/hidden from user? Replication strategy is pluggable - opt in key space by key space ? > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15972042#comment-15972042 ] Jonathan Ellis commented on CASSANDRA-13442: Man, this makes me really nervous. Reducing replication automagically is ripping the safety guard off a chainsaw and handing it to a ten year old. Remember that those replicas aren't just for consistency but in case your hardware fails: if you have three copies and you lose one, no big deal, you still have two to restore from. Just two copies? If anything goes wrong with that other copy while you repair you are SOL. Optimizing away redundant queries a la 7168? Sign me up. But I think removing that "redundant" data and making RF not actually mean RF is going too far. > The topology of the cluster would also have a new dimension that the drivers > would need to consider. Since for CL.ONE queries you would need to only use > one of the replicas with all the data on it. Yes, I think there are going to be multiple places where this gets more complicated than it looks at first. > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15972014#comment-15972014 ] T Jake Luciani commented on CASSANDRA-13442: .bq This ticket is not about changing consistency levels and doesn't require applications to change their usage of consistency levels to benefit. 7168 added a new CL as a way to opt-in to this new feature. Once its fully vetted it would be trivial to make it automatically use it when appropriate. bq. 7168 also does not have reducing storage requirements as a goal. This idea seem risky to me. At the least it should be opt-in. From a operators perspective you would need to consider how to handle bootstrapping or replacing a node. Also, how to handle backup and restores, etc. The topology of the cluster would also have a new dimension that the drivers would need to consider. Since for CL.ONE queries you would need to only use one of the replicas with all the data on it. > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15971985#comment-15971985 ] Ariel Weisberg commented on CASSANDRA-13442: CASSANDRA-7168 is about adding a consistency level. This ticket is not about changing consistency levels and doesn't require applications to change their usage of consistency levels to benefit. 7168 also does not have reducing storage requirements as a goal. They both share the observation that repair creates opportunities to make assumptions about the data set and what is necessary to provide strong consistency. > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15971975#comment-15971975 ] T Jake Luciani commented on CASSANDRA-13442: How is this not a duplicate of CASSANDRA-7168 > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15971858#comment-15971858 ] sankalp kohli commented on CASSANDRA-13442: --- cc [~jbellis] What do you think about this idea? > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (CASSANDRA-13442) Support a means of strongly consistent highly available replication with storage requirements approximating RF=2
[ https://issues.apache.org/jira/browse/CASSANDRA-13442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15966376#comment-15966376 ] Jeff Jirsa commented on CASSANDRA-13442: Potentially related background reading: CASSANDRA-7168 > Support a means of strongly consistent highly available replication with > storage requirements approximating RF=2 > > > Key: CASSANDRA-13442 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13442 > Project: Cassandra > Issue Type: Improvement > Components: Compaction, Coordination, Distributed Metadata, Local > Write-Read Paths >Reporter: Ariel Weisberg > > Replication factors like RF=2 can't provide strong consistency and > availability because if a single node is lost it's impossible to reach a > quorum of replicas. Stepping up to RF=3 will allow you to lose a node and > still achieve quorum for reads and writes, but requires committing additional > storage. > The requirement of a quorum for writes/reads doesn't seem to be something > that can be relaxed without additional constraints on queries, but it seems > like it should be possible to relax the requirement that 3 full copies of the > entire data set are kept. What is actually required is a covering data set > for the range and we should be able to achieve a covering data set and high > availability without having three full copies. > After a repair we know that some subset of the data set is fully replicated. > At that point we don't have to read from a quorum of nodes for the repaired > data. It is sufficient to read from a single node for the repaired data and a > quorum of nodes for the unrepaired data. > One way to exploit this would be to have N replicas, say the last N replicas > (where N varies with RF) in the preference list, delete all repaired data > after a repair completes. Subsequent quorum reads will be able to retrieve > the repaired data from any of the two full replicas and the unrepaired data > from a quorum read of any replica including the "transient" replicas. -- This message was sent by Atlassian JIRA (v6.3.15#6346)
