You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Sean Busbey (Jira)" <ji...@apache.org> on 2021/06/07 14:03:00 UTC
[jira] [Updated] (HBASE-25973) Balancer should explain progress in
a better way in log
[ https://issues.apache.org/jira/browse/HBASE-25973?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Sean Busbey updated HBASE-25973:
--------------------------------
Status: Patch Available (was: Open)
> Balancer should explain progress in a better way in log
> -------------------------------------------------------
>
> Key: HBASE-25973
> URL: https://issues.apache.org/jira/browse/HBASE-25973
> Project: HBase
> Issue Type: Bug
> Components: Balancer
> Affects Versions: 3.0.0-alpha-1
> Reporter: Clara Xiong
> Assignee: Clara Xiong
> Priority: Major
>
> In the log, balancer logs at info level at the beginning of run:
>
> balancer.StochasticLoadBalancer: start StochasticLoadBalancer.balancer, initCost=277.3479243125063, functionCost=RegionCountSkewCostFunction : (500.0, 0.3749771215224234); ServerLocalityCostFunction : (25.0, 0.5807483226644186); RackLocalityCostFunction : (15.0, 0.0); TableSkewCostFunction : (1000.0, 0.0019704142954972883); StoreFileCostFunction : (200.0, 0.3668512059459341); computedMaxSteps: 42270438200
> the cost is reported without context, it is hard for operator to understand how unbalanced the cluster is for balancer and how much progress we are making.
> For a large cluster, the calculation can take a long time, we also need to let operator understand that it will take up to the max time to complete the calculation.
> At the end of computation:
> balancer.StochasticLoadBalancer: Finished computing new load balance plan. Computation took PT40M0.006S to try 1036409 different iterations. Found a solution that moves 161926 regions; Going from a computed cost of 118.75715593924485 to a new cost of 1.5509126920967042
> The time to compute the plan is also printed in a format that is not human readable. we also need to let operator understand that balancer is just submitting the plan and it be up to execution to complete the move.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)