Harsh J created HBASE-8020:
------------------------------
Summary: CompactionRequest should not be based on regionId, but
the encoded name instead
Key: HBASE-8020
URL: https://issues.apache.org/jira/browse/HBASE-8020
Project: HBase
Issue Type: Bug
Affects Versions: 0.94.2
Reporter: Harsh J
Priority: Minor
The compaction state check call of HRegionServer#getCompactionState extracts
out the regionId (timestamp) of a region to store/check the compaction state.
In usual runtime this is fine, as the timestamp is granular and varied for each
region, making it practically unique but in no way should a timestamp be
considered unique. A bug that comes out of this semi-bad assumption is that
when a user pre-splits their table, the regionId is all the same across all
regions created, leading to this check call returning improper values.
The ideal identifier could be the table name + encoded region name, if not the
whole HRegionInfo identifier string itself.
Given that the user API just sends a normal identifier which is reevaluated,
such a change shouldn't break anything.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
