hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Duo Zhang (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-19812) TestFlushSnapshotFromClient fails because of failing region.flush
Date Wed, 17 Jan 2018 22:12:00 GMT

    [ https://issues.apache.org/jira/browse/HBASE-19812?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16329576#comment-16329576

Duo Zhang commented on HBASE-19812:

I think this could be a general reason that causes lots of flakey tests since we usually expected
a file on disk when we call region.flush. So I think we should try to fix the general problem?

> TestFlushSnapshotFromClient fails because of failing region.flush
> -----------------------------------------------------------------
>                 Key: HBASE-19812
>                 URL: https://issues.apache.org/jira/browse/HBASE-19812
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Duo Zhang
>            Priority: Major
>         Attachments: 19812.patch, 19812.patch
> {noformat}
> 2018-01-17 06:43:48,390 INFO  [MemStoreFlusher.1] regionserver.HRegion(2516): Flushing
1/1 column families, memstore=549.25 KB
> 2018-01-17 06:43:48,390 DEBUG [MemStoreFlusher.1] regionserver.CompactingMemStore(205):
FLUSHING TO DISK: region test,5,1516171425662.acafc22e1f8132285eae5362d0df536a.store: fam
> 2018-01-17 06:43:48,406 DEBUG [RpcServer.default.FPBQ.Fifo.handler=4,queue=0,port=42601-inmemoryCompactions-1516171428312]
regionserver.CompactionPipeline(206): Compaction pipeline segment Type=CSLMImmutableSegment,
empty=no, cellCount=17576, cellSize=562432, totalHeapSize=1828120, min timestamp=1516171428258,
max timestamp=1516171428258Num uniques -1;  flattened
> 2018-01-17 06:43:48,406 DEBUG [MemStoreFlusher.1] regionserver.CompactionPipeline(128):
Swapping pipeline suffix; before=1, new segement=null
> 2018-01-17 06:43:48,455 DEBUG [Time-limited test] regionserver.HRegion(2201): NOT flushing
memstore for region test,5,1516171425662.acafc22e1f8132285eae5362d0df536a., flushing=true,
> {noformat}
> You can see that we start a background flush first, and then we decided to do an in memory
compaction, at the same time we call the region.flush from test, and it find that the region
is already flushing so it give up.
> This test is a bit awkward that we create the table with 6 regions which start key is
0,1,2,3,4,5, but when loading data we use 'aaa' to 'zzz', so there is only one region has
data. And in the above scenario the only one region gives up flushing, then there is no data,
and then our test fails.

This message was sent by Atlassian JIRA

View raw message