[
https://issues.apache.org/jira/browse/HBASE-17039?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15649678#comment-15649678
]
Charlie Qiangeng Xu commented on HBASE-17039:
---------------------------------------------
Hi Enis, The reason that we don't use Stochastic one is because we used to find
the regions were moved erratically.
And the conf parameter that control the StochasticLB is a bit hard to
fine-tune.(the chance that we can play back and forth on large scale production
cluster is limited)
Another factor that might contribute to the decision would be that
simpleBalancer can guarantee table level region balance, which is very crucial
in our use case.
Those are old conclusions though, may be outdated, since StochasticLB has been
updated by several patches after
we observe those disadvantages.
We will definitely retry StochasticLB after these busy months and will give you
feed back after that :)
> SimpleLoadBalancer schedules large amount of invalid region moves
> -----------------------------------------------------------------
>
> Key: HBASE-17039
> URL: https://issues.apache.org/jira/browse/HBASE-17039
> Project: HBase
> Issue Type: Bug
> Components: Balancer
> Affects Versions: 2.0.0, 1.2.3, 1.1.7
> Reporter: Charlie Qiangeng Xu
> Assignee: Charlie Qiangeng Xu
> Attachments: HBASE-17039.patch
>
>
> After increasing one of our clusters to 1600 nodes, we observed a large
> amount of invalid region moves(more than 30k moves) fired by the balance
> chore. Thus we simulated the problem and printed out the balance plan, only
> to find out many servers that had two regions for a certain table(we use by
> table strategy), sent out both regions to other two servers that have zero
> region.
> In the SimpleLoadBalancer's balanceCluster function,
> the code block that determines the underLoadedServers might have a problem:
> {code}
> if (load >= min && load > 0) {
> continue; // look for other servers which haven't reached min
> }
> int regionsToPut = min - load;
> if (regionsToPut == 0)
> {
> regionsToPut = 1;
> }
> {code}
> if min is zero, some server that has load of zero, which equals to min would
> be marked as underloaded, which would cause the phenomenon mentioned above.
> Since we increased the cluster's size to 1600+, many tables that only have
> 1000 regions, now would encounter such issue.
> By fixing it up, the balance plan went back to normal.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
