[
https://issues.apache.org/jira/browse/HBASE-3530?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12994399#comment-12994399
]
Jonathan Gray commented on HBASE-3530:
--------------------------------------
In some of the HLog improvements that Dhruba and I have been working on, we've
become dependent on the LogSyncer thread. However, this is also still using an
HBase-side group commit. Just an FYI (alternative implementations of HLog
might be a neat idea, there are lots of different ways to optimize depending on
the use case)... don't let it hold you up.
Expect some patches this week around this stuff. Currently tracking down a
pesky race condition.
> Do not start LogSyncer when not in deferred flush mode
> ------------------------------------------------------
>
> Key: HBASE-3530
> URL: https://issues.apache.org/jira/browse/HBASE-3530
> Project: HBase
> Issue Type: Improvement
> Components: regionserver
> Affects Versions: 0.90.0
> Reporter: Lars George
> Priority: Minor
> Fix For: 0.92.0
>
>
> Currently we always start the LogSyncer thread, even when not necessary
> because the "deferred flush" is set to false and the sync() is called on
> every record. We could disable the thread and not have it flush every n secs
> unnecessarily.
--
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
