[
https://issues.apache.org/jira/browse/HBASE-16301?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15400389#comment-15400389
]
huaxiang sun commented on HBASE-16301:
--------------------------------------
Thanks [~enis] for review. I thought about moving it inside
isTooManyStoreFiles() when I first worked on the patch. But at that time, I
thought that isTooManyStoreFiles() really means too many store files so aborted
that idea. Looking at flushOneForGlobalPressure() again, it seems that it
should be moved to isTooManyStoreFiles() to make the logic right as it really
means flushable.
One minor improvement here is that for bestFlushableRegion and bestAnyRegion,
they can be resolved in one loop and return a pair instead of two separate
loops.
https://github.com/apache/hbase/blob/master/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/MemStoreFlusher.java#L153
{code}
Region bestFlushableRegion = getBiggestMemstoreRegion(regionsBySize,
excludedRegions, true);
// Find the biggest region, total, even if it might have too many flushes.
Region bestAnyRegion = getBiggestMemstoreRegion(
regionsBySize, excludedRegions, false);
{code}
> Trigger flush without waiting when compaction is disabled on a table
> --------------------------------------------------------------------
>
> Key: HBASE-16301
> URL: https://issues.apache.org/jira/browse/HBASE-16301
> Project: HBase
> Issue Type: Bug
> Affects Versions: 2.0.0
> Reporter: huaxiang sun
> Assignee: huaxiang sun
> Priority: Minor
> Attachments: HBASE-16301-v001.patch
>
>
> When compaction is disabled on a table, flush needs to wait
> MemStoreFlusher#blockingWaitTime (default value is 90 seconds) before it goes
> ahead to flush. It has side effect that client may be blocked due to
> RegionTooBusyException. Please see the mail sent to dev list.
> http://mail-archives.apache.org/mod_mbox/hbase-dev/201607.mbox/%[email protected]%3E
> I guess that the right behavior is to do flush without waiting if compaction
> is disabled on a table. Attached a patch.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
