[
https://issues.apache.org/jira/browse/HBASE-16579?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15472185#comment-15472185
]
Ted Yu commented on HBASE-16579:
--------------------------------
The concurrency comes from obtaining exclusive lock on region(s) of
hbase:backup table.
e.g. one region corresponds to one rsgroup.
If one session fails, it may not be necessary to roll whole hbase:backup table
to the snapshot.
> Concurrent backup / restore support
> -----------------------------------
>
> Key: HBASE-16579
> URL: https://issues.apache.org/jira/browse/HBASE-16579
> Project: HBase
> Issue Type: Improvement
> Reporter: Ted Yu
> Labels: backup
>
> Currently we take exclusive lock on hbase:backup table in backup / restore
> procedures.
> There is only one backup or restore operation which can run at given time.
> We should consider using shared lock on hbase:backup table in the procedures
> so that concurrent backup / restore operations can proceed.
> The concurrency is useful when table sets don't overlap.
> We may devise other mechanism(s) when there is overlap among the table sets.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
