[
https://issues.apache.org/jira/browse/HBASE-22596?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16867036#comment-16867036
]
Zach York commented on HBASE-22596:
-----------------------------------
The other consideration is that looking at the code it appears all of them
still share the same core service with only one thread. So even if you change
the wake threshold it appears that it would be possible for one of these chores
to monopolize the time.
> [Chore] Separate the execution period between CompactionChecker and
> PeriodicMemStoreFlusher
> -------------------------------------------------------------------------------------------
>
> Key: HBASE-22596
> URL: https://issues.apache.org/jira/browse/HBASE-22596
> Project: HBase
> Issue Type: Improvement
> Components: Compaction
> Reporter: Reid Chan
> Assignee: Reid Chan
> Priority: Minor
> Attachments: HBASE-22596.master.001.patch
>
>
> The story started at tuning RS performance where I found:
> {code}
> public static final String THREAD_WAKE_FREQUENCY =
> "hbase.server.thread.wakefrequency";
> ...
> this.threadWakeFrequency = conf.getInt(HConstants.THREAD_WAKE_FREQUENCY, 10 *
> 1000);
> ...
> this.compactionChecker = new CompactionChecker(this,
> this.threadWakeFrequency, this);
> this.periodicFlusher = new PeriodicMemStoreFlusher(this.threadWakeFrequency,
> this);
> {code}
> CompactionChecker and PeriodicMemStoreFlusher execution period are bound
> together. (as well as LeaseChecker)
> This issue is going to introduce two new parameters such that user/admin can
> tune them according to business workload.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
