[
https://issues.apache.org/jira/browse/HBASE-19121?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16310665#comment-16310665
]
stack commented on HBASE-19121:
-------------------------------
Here is a case we need tooling for: region is reporting the below in the Master:
2017-12-31 10:12:52,611 WARN [ProcExecTimeout] assignment.AssignmentManager:
TODO Handle stuck in transition: rit=OPENING,
location=node8.com,16020,1514693333206, table=work_proposed,
region=5b4b9a7b4e58da39a2072fdcb512df2f
We should NEVER arrive at the above state. Up to this we've been trying to plug
all the holes in AMv2 that might provoke the above but it is unlikely we'll see
all failure scenarios before we ship 2.0.0 so there will be cases where regions
will end up in the unassignable state (and operator will see logs like those
above with associated regions offline). An hbck2 would intervene and if the
region unassigned ten minutes or more, it'd move region back to an assignable
state via Master invocations.
See the thread [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is
available" for more. See [~jmspaggi] notes.
> HBCK for AMv2
> -------------
>
> Key: HBASE-19121
> URL: https://issues.apache.org/jira/browse/HBASE-19121
> Project: HBase
> Issue Type: Bug
> Components: hbck
> Reporter: stack
> Fix For: 2.0.0-beta-2
>
>
> We don't have an hbck for the new AM. Old hbck may actually do damage going
> against AMv2.
> Fix.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
