[
https://issues.apache.org/jira/browse/HBASE-19384?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16273836#comment-16273836
]
Rajeshbabu Chintaguntla commented on HBASE-19384:
-------------------------------------------------
[~elserj]
bq. Rajeshbabu Chintaguntla any chance you could share the specifics on how the
coprocessors are set up in which you see this? Or, if you really want to go the
extra mile, distill it down to a generic test case?
Sure will write test case for this.
[~stack]
bq. So you have multiple coprocessors stacked on a region. One intercepts
preAppend (or preIncrement) to return its own result instead. Are you saying
that this result is overwritten by the null that subsequent coprocessors return?
Correct.
bq. Is the problem our removal of 'complete'? i.e. HBASE-19123 Purge 'complete'
support from Coprocesor Observers ? Thanks.
Yes it's because of removal of complete so we are not able to skip running
subsequent coprocessors which do not have any implementation for preAppend or
preIncrement hooks.
> Results returned by preAppend hook in a coprocessor are replaced with null
> from other coprocessor even on bypass
> ----------------------------------------------------------------------------------------------------------------
>
> Key: HBASE-19384
> URL: https://issues.apache.org/jira/browse/HBASE-19384
> Project: HBase
> Issue Type: Bug
> Components: Coprocessors
> Reporter: Rajeshbabu Chintaguntla
> Assignee: Rajeshbabu Chintaguntla
> Priority: Critical
> Fix For: 2.0.0-beta-1
>
>
> Phoenix adding multiple coprocessors for a table and one of them has
> preAppend and preIncrement implementation and bypass the operations by
> returning the results. But the other coprocessors which doesn't have any
> implementation returning null and the results returned by previous
> coprocessor are override by null and always going with default implementation
> of append and increment operations. But it's not the case with old versions
> and works fine on bypass.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
