Anoop Sam John commented on HBASE-16608:

As Ram reported that he can see so many GC activity and even many full GCs, I 
repeated the tests with same config. We have 30 GB heap space and 42% as global 
memstore size.  The chunk pool is configured with 100% as max size. Means whole 
42% of this size can be cached now.  So ideally we should be getting chunk from 
pool always.  I confirmed that with DefaultMemstore case..  Doing the same load 
test with CompactingMemstore I can see so many times we miss a getChunk call on 
MSLABPool..  Seems the bug fix that was added in latest patch is not releasing 
chunks correctly. This makes us to create on demand chunks so many times and 
there is all chances that they get promoted to old gen and so more chances of 
even full GCs.  Just reporting back what is the actual issue.

> Introducing the ability to merge ImmutableSegments without copy-compaction or 
> SQM usage
> ---------------------------------------------------------------------------------------
>                 Key: HBASE-16608
>                 URL: https://issues.apache.org/jira/browse/HBASE-16608
>             Project: HBase
>          Issue Type: Sub-task
>            Reporter: Anastasia Braginsky
>            Assignee: Anastasia Braginsky
>         Attachments: HBASE-16417-V02.patch, HBASE-16417-V04.patch, 
> HBASE-16417-V06.patch, HBASE-16417-V07.patch, HBASE-16417-V08.patch, 
> HBASE-16417-V10.patch, HBASE-16608-V01.patch, HBASE-16608-V03.patch, 
> HBASE-16608-V04.patch, HBASE-16608-V08.patch

This message was sent by Atlassian JIRA

Reply via email to