[
https://issues.apache.org/jira/browse/HBASE-28841?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Wellington Chevreuil resolved HBASE-28841.
------------------------------------------
Resolution: Fixed
Merged to master, branch-3 and branch-2. Thanks for the contribution,
[~janardhan.hungund]!
> Modify default value of hbase.bucketcache.persistence.chunksize to 10K
> ----------------------------------------------------------------------
>
> Key: HBASE-28841
> URL: https://issues.apache.org/jira/browse/HBASE-28841
> Project: HBase
> Issue Type: Bug
> Affects Versions: 3.0.0-beta-1, 4.0.0-alpha-1, 2.7.0
> Reporter: Janardhan Hungund
> Assignee: Janardhan Hungund
> Priority: Major
> Labels: pull-request-available
>
> Currently, the default value of the configuration parameter
> "hbase.bucketcache.persistence.chunksize" is 10 million (10000000).
> This is the number of block entries that are processed during the persistence
> of bucket-cache backing map to the persistence file. During the testing, it
> was found that, this high number of chunksize resulted in high heap
> utilisation in region servers leading to longer GC pauses which also led to
> server crashes intermittently.
> When the value of this configuration is set to 10K(10000), the cache remains
> stable. No GC delays are observed. Also no server crashes are observed.
> The jmap outputs collected against the regionservers showed reduced memory
> utilisation from 4.5-5GB to 1-1.5GB for the objects related to persistence
> code.
> Hence, we need to adjust the default value of this configuration parameter to
> 10K.
> Thanks,
> Janardhan
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
