[
https://issues.apache.org/jira/browse/HBASE-7965?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13591241#comment-13591241
]
Enis Soztutar commented on HBASE-7965:
--------------------------------------
bq. At some point the master will be a version that expects the locks to be
honored. The region servers will be in some indeterminate state.
If you are doing rolling upgrade from a version which does not have table
locks, to one that have 0.94.5 -> 0.95.7, and enale table locks, when master is
updated, all ops will start acquiring the lock. And when regionservers are
upgraded, the splits will start to get read locks. Until all of the cluster is
upgraded, the situation will be the same as w/o table locks, that concurrent
table ops and regions splits will cause headaches. It will be ok, if some of
the servers knows table locks, while the master / other RSs does not know. You
wont get the exclusion guarantee though.
> Port table locking to 0.94 (HBASE-7305, HBASE-7546, HBASE-7933)
> ---------------------------------------------------------------
>
> Key: HBASE-7965
> URL: https://issues.apache.org/jira/browse/HBASE-7965
> Project: HBase
> Issue Type: New Feature
> Components: master, Zookeeper
> Reporter: Enis Soztutar
> Assignee: Enis Soztutar
> Fix For: 0.94.7
>
>
> Port table locking to 0.94 (HBASE-7305, HBASE-7546, HBASE-7933). This is a
> new feature, but there has been some interest, and it is necessary for
> snapshots, and online merge, which are also candidates for backport.
> If we port snapshots, we might need HBASE-7848 as well.
> We can also do disabled by default.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
