[
https://issues.apache.org/jira/browse/HBASE-8459?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jonathan Hsieh updated HBASE-8459:
----------------------------------
Description:
Currently, online snapshots assign work by region server. This works when
regionseveers do not fail or moved regions. In 0.95+, the table locks prevent
moves while snapshotting is in progress but a failed RS during a snapshot may
still fail in the verification step. This is because the failure would trigger
an assignment which may occur after the work for the regionserver is
determined, leaving some regions unsnapshotted, thus leaving it to fail in the
verification step.
If we instead assigned snapshotting work by region name, a moved region could
be snapshotted by the newly assigned region server.
was:
Currently, online snapshots assign work by region server. This works when
regionseveers do not fail or moved regions. In 0.95+, the table locks prevent
moves while snapshotting is in progress but a failed RS during a snapshot may
still fail in the verification step. This is because the failure would trigger
an assignment which may occur after the work for the regionserver is
determined, leaving some regions unsnapshotted, thus leaving it to fail in the
verification step.
If we instead assigned snapshotting work by region name, a moved region could
be snapshotted by the newly assigned region.
> make online region snapshotting based on region names instead of region
> server names
> ------------------------------------------------------------------------------------
>
> Key: HBASE-8459
> URL: https://issues.apache.org/jira/browse/HBASE-8459
> Project: HBase
> Issue Type: Bug
> Components: snapshots
> Affects Versions: 0.94.6.1, 0.95.0
> Reporter: Jonathan Hsieh
>
> Currently, online snapshots assign work by region server. This works when
> regionseveers do not fail or moved regions. In 0.95+, the table locks
> prevent moves while snapshotting is in progress but a failed RS during a
> snapshot may still fail in the verification step. This is because the
> failure would trigger an assignment which may occur after the work for the
> regionserver is determined, leaving some regions unsnapshotted, thus leaving
> it to fail in the verification step.
> If we instead assigned snapshotting work by region name, a moved region could
> be snapshotted by the newly assigned region server.
--
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
