[
https://issues.apache.org/jira/browse/HBASE-23265?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ondrej Kvasnicka updated HBASE-23265:
-------------------------------------
Description:
According to our observation, a running coprocessor that is automatically
stopped prior to region split will not be automatically restarted for the
region that was about to be split in case the split attempt is rolled back.
The expected behavior would be for the coprocessor to be automatically
restarted in such situation.
According to [this
conversation|https://lists.apache.org/thread.html/9f83d32c50e0f9b05fff183897f5a3ccdb3c2e94b0e9d1a0f9064646@%3Cuser.hbase.apache.org%3E]
the lack of the automatic restart looks more like an omission rather than an
intended behavior.
A possible work-around seems to be handling the restart explicitly in
RegionObserver#postRollBackSplit() implementation, such as in [this
example|https://github.com/splicemachine/spliceengine/blob/aa9d66e3d33ea04d67203eb86b3e6e3ba768e267/hbase_pipeline/src/main/java/com/splicemachine/derby/hbase/SpliceIndexObserver.java#L206].
was:
According to our observation, a running coprocessor that is automatically
stopped prior to region split will not be automatically restarted for the
region that was about to be split in case the split attempt is rolled back.
The expected behavior would be for the coprocessor to be automatically
restarted in such situation.
According to [this
conversation|https://lists.apache.org/thread.html/9b29afc5aa4994b67bc29a712cef276c7ff10d453f820050fdb760c8@%3Cuser.hbase.apache.org%3E]
the lack of the automatic restart looks more like an omission rather than an
intended behavior.
A possible work-around seems to be handling the restart explicitly in
RegionObserver#postRollBackSplit() implementation, such as in [this
example|https://github.com/splicemachine/spliceengine/blob/aa9d66e3d33ea04d67203eb86b3e6e3ba768e267/hbase_pipeline/src/main/java/com/splicemachine/derby/hbase/SpliceIndexObserver.java#L206].
> Coprocessor restart after region split rollback
> -----------------------------------------------
>
> Key: HBASE-23265
> URL: https://issues.apache.org/jira/browse/HBASE-23265
> Project: HBase
> Issue Type: Improvement
> Components: Coprocessors
> Affects Versions: 1.1.2
> Reporter: Ondrej Kvasnicka
> Priority: Minor
>
> According to our observation, a running coprocessor that is automatically
> stopped prior to region split will not be automatically restarted for the
> region that was about to be split in case the split attempt is rolled back.
> The expected behavior would be for the coprocessor to be automatically
> restarted in such situation.
> According to [this
> conversation|https://lists.apache.org/thread.html/9f83d32c50e0f9b05fff183897f5a3ccdb3c2e94b0e9d1a0f9064646@%3Cuser.hbase.apache.org%3E]
> the lack of the automatic restart looks more like an omission rather than an
> intended behavior.
> A possible work-around seems to be handling the restart explicitly in
> RegionObserver#postRollBackSplit() implementation, such as in [this
> example|https://github.com/splicemachine/spliceengine/blob/aa9d66e3d33ea04d67203eb86b3e6e3ba768e267/hbase_pipeline/src/main/java/com/splicemachine/derby/hbase/SpliceIndexObserver.java#L206].
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
