[
https://issues.apache.org/jira/browse/HBASE-3583?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13001722#comment-13001722
]
Mingjie Lai commented on HBASE-3583:
------------------------------------
After reviewed the current implementation, I think it makes more sense to pull
the scannerNext and scannerClose hooks from HRegion to HRegionServer, instead
of using a flag to indicate whether to invoke coprocessor hooks or not. Reason
1) the internalScanner are called at several different places, there are
potential misfires as this case; 2) it would be ugly to pass a boolean flag to
the internalscanner(it's an interface).
Patch posted here and reviewboard.
https://review.cloudera.org/r/1617/
> Coprocessors: RegionObserver: ScannerNext and ScannerClose hooks are called
> when get() is invoked
> -------------------------------------------------------------------------------------------------
>
> Key: HBASE-3583
> URL: https://issues.apache.org/jira/browse/HBASE-3583
> Project: HBase
> Issue Type: Bug
> Components: coprocessors
> Affects Versions: 0.92.0
> Reporter: Mingjie Lai
> Assignee: Mingjie Lai
> Fix For: 0.92.0
>
>
> RegionObserver upcalls are expected to be triggered by corresponding client
> calls.
> I found that if a HTable.get() is issued, ScannerNext, and ScannerClose hooks
> are also invoked.
> Here is the reason: HRegion.get() is implemented with an internal scanner:
> {code}
> InternalScanner scanner = null;
> try {
> scanner = getScanner(scan);
> scanner.next(results);
> } finally {
> if (scanner != null)
> scanner.close();
> }
> {code}
> where scanner.next, and scanner.close() are implemented with RegionObserver
> hooks.
--
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
