[
https://issues.apache.org/jira/browse/HBASE-24465?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Nick Dimiduk resolved HBASE-24465.
----------------------------------
Resolution: Not A Problem
I think this is this actually a problem. The total size of a generated merge
action is limited to the average region size. If average region size exceeds
max region file size, region servers will split their regions, thus driving
down the average.
> Normalizer should consider region max file size when planning merges
> --------------------------------------------------------------------
>
> Key: HBASE-24465
> URL: https://issues.apache.org/jira/browse/HBASE-24465
> Project: HBase
> Issue Type: Improvement
> Components: master, Normalizer
> Reporter: Nick Dimiduk
> Priority: Major
>
> When the normalizer plans its actions, it does not consider
> {{hbase.hregion.max.filesize}}. This means it could get into a merge/split
> loop in collaboration with the region server: the normalizer sees two regions
> are smaller than the table average, it merges them. The resulting region is
> larger that this max file size, so the region server splits them. Repeat.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
