[
https://issues.apache.org/jira/browse/HBASE-13560?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14566281#comment-14566281
]
Ted Yu edited comment on HBASE-13560 at 5/31/15 1:05 AM:
---------------------------------------------------------
{code}
497 //prevent large compaction thread pool stealing job from small
compaction queue.
498 cst.shutdownLongCompactions();
{code}
Can you explain the above change a little more ? In production, large
compaction thread pool would always be running (as long as region server is
running), right ?
was (Author: [email protected]):
{code}
497 //prevent large compaction thread pool stealing job from small
compaction queue.
498 cst.shutdownLongCompactions();
{code}
Can you explain the above change a little more ? In production, large
compaction thread pool would always be running, right ?
> Large compaction queue should steal from small compaction queue when idle
> -------------------------------------------------------------------------
>
> Key: HBASE-13560
> URL: https://issues.apache.org/jira/browse/HBASE-13560
> Project: HBase
> Issue Type: Bug
> Components: Compaction
> Affects Versions: 2.0.0
> Reporter: Elliott Clark
> Assignee: Changgeng Li
> Attachments: queuestealwork-v1.patch, queuestealwork-v4.patch,
> queuestealwork-v5.patch, queuestealwork-v6.patch, queuestealwork-v7.patch
>
>
> If you tune compaction threads so that a server is never over commited when
> large and small compaction threads are busy then it should be possible to
> have the large compactions steal work.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
