[
https://issues.apache.org/jira/browse/HBASE-3332?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jonathan Gray updated HBASE-3332:
---------------------------------
Attachment: HBASE-3332-v1.patch
Previously ServerShutdownHandler would not handle any regions on the dead
server if they were already in transition. However, in the TimeoutMonitor, we
will never force a region off of a server when it's PENDING_CLOSE or CLOSING
because the expectation was we would handle it in server expiration. But
again, over there, we did nothing to all RIT.
Change makes it so we will force reassignment of RIT if PENDING_CLOSE or
CLOSING.
Patch is not done yet, I need to make sure master won't screw something up and
I want to add a test.
> Regions stuck in transition after RS failure
> --------------------------------------------
>
> Key: HBASE-3332
> URL: https://issues.apache.org/jira/browse/HBASE-3332
> Project: HBase
> Issue Type: Bug
> Affects Versions: 0.90.0
> Reporter: Todd Lipcon
> Assignee: Jonathan Gray
> Priority: Blocker
> Attachments: HBASE-3332-v1.patch, log.txt
>
>
> Testing 0.90rc2 I ran into this issue. The test scenario was to kill -9 the
> server hosting ROOT and META, and before it had been detected, run "balancer"
> from the shell. After logs were split and regions were reassigned, I ended up
> with some regions stuck in transition.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
