[
https://issues.apache.org/jira/browse/HBASE-21875?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16766964#comment-16766964
]
Duo Zhang commented on HBASE-21875:
-----------------------------------
We have an UncaughtExceptionHandler for the thread pool in
RSProcedureDispatcher but it seems that it never log anything, but I'm sure my
assertion in the RsExecutor in the TestAssignmentManager is broken, as if I
added the catch Throwable in the call method I can output the error, but if I
removed the catch in call method, the test just hang there...
Let me dig more. This maybe the root cause that why some OpenRegionProcedure
can hang there forever, we have some bugs in the OpenRegionProcedure
implementation and introduce a RuntimeException, but there is no warning log so
it looks like the OpenRegionProcedure is hang there forever...
> Change the retry logic in RSProcedureDispatcher to 'retry by default, only if
> xxx'
> ----------------------------------------------------------------------------------
>
> Key: HBASE-21875
> URL: https://issues.apache.org/jira/browse/HBASE-21875
> Project: HBase
> Issue Type: Improvement
> Components: proc-v2
> Reporter: Duo Zhang
> Assignee: Duo Zhang
> Priority: Major
> Fix For: 3.0.0, 2.2.0, 2.3.0
>
> Attachments: HBASE-21875-v1.patch, HBASE-21875-v2.patch,
> HBASE-21875.patch
>
>
> For now it is not retry by default, only if xxx.
> In executeProcedures, we will only throw a fixed set of exception, so we
> should change to retry by default, and check for the exceptions which we do
> not need to retry.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
