[jira] [Commented] (HBASE-15490) Two CompactionThroughputControllerFactory co-exist in branch-1

["Yu Li (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-15490?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15202905#comment-15202905 ] 

Yu Li commented on HBASE-15490:
-------------------------------

Checking current branch-1 codes, the to-be-removed {{CompactionThroughputControllerFactory}} in o.a.h.h.regionserver.compactions is only called by our UT case, and only the new one in o.a.h.h.regionserver.throttle will take effect in the real world. Since 1.1.3 is released where {{PressureAwareCompactionThroughputController}} is set as default with 20MB/s as throttle threshold, I'm wondering whether we should change the default back to {{NoLimitThroughputController}} or add a release note in HBASE-14969 identifying the change (probably need a discussion on the default lower/upper bound also).

Please let me know your thoughts guys. Thanks.

> Two CompactionThroughputControllerFactory co-exist in branch-1
> --------------------------------------------------------------
>
>                 Key: HBASE-15490
>                 URL: https://issues.apache.org/jira/browse/HBASE-15490
>             Project: HBase
>          Issue Type: Bug
>    Affects Versions: 1.1.3
>            Reporter: Yu Li
>            Assignee: Yu Li
>
> Currently there're two {{CompactionThroughputControllerFactory}} in our branch-1 code base (one in {{o.a.h.h.regionserver.compactions}} package, the other in {{o.a.h.h.regionserver.throttle}}) and both are in use.
> This is a regression of HBASE-14969 and only exists in branch-1. We should remove the one  in {{o.a.h.h.regionserver.compactions}}, and change the default compaction throughput controller back to {{NoLimitThroughputController}} to keep compatible with previous branch-1 version
> Thanks [~ghelmling] for pointing out the issue.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)