[
https://issues.apache.org/jira/browse/HBASE-22929?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16916525#comment-16916525
]
Yechao Chen edited comment on HBASE-22929 at 8/27/19 9:07 AM:
--------------------------------------------------------------
It seems :
the chunkIdMap of the ChunkCreator increment incorrect
the chunk size default is 2MB,global memstore limit is 32GB*0.4
the maxCount of the chunks should be (32GB*0.4)/2MB = 6553
the chunkIdMap size should be smaller than 6553
but the heap dump shows the size of chunkIdMap has 14579 entries
was (Author: chenyechao):
It seems :
the chunkIdMap of the ChunkCreator increament incorrect
the maxCount of the chunks should be (32GB*0.4)/2MB = 6553
the chunkIdMap size should be smaller than 6553
but the heap dump shows the size of chunkIdMap has 14579 entries
> MemStoreLAB ChunkCreator may memory leak
> -----------------------------------------
>
> Key: HBASE-22929
> URL: https://issues.apache.org/jira/browse/HBASE-22929
> Project: HBase
> Issue Type: Bug
> Affects Versions: 2.1.2
> Reporter: Yechao Chen
> Priority: Major
> Attachments: OOM_log.png, hbase-site.xml, hbase_heap_monitor.png,
> hbase_rs_heap_dump_mat_1.png,
> hbase_rs_heap_dump_mat_ChunkCreator_chunkIdMap.png, hbase_rs_mem_used.png
>
>
> We use hbase 2.1.2 with memstorelab enable
> RegionServer crashed case of oom
> I dump the heap ,found the ChunkCreator may be memory leak
> The heap is 32GB,
> hbase.regionserver.global.memstore.size=0.4,
> hbase.hregion.memstore.mslab.enabled=true
> hbase.hregion.memstore.chunkpool.initialsize=0.5,
> hbase.hregion.memstore.chunkpool.maxsize=1.0
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
