[
https://issues.apache.org/jira/browse/HBASE-6180?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13396047#comment-13396047
]
Jesse Yates commented on HBASE-6180:
------------------------------------
Further recommendation for the timestamp based, swapping memstore/flush
approach - distributed splits cause the WALs to be moved to another directory,
causing our symlinks (via the Reference class) to become invalid. This isn't a
problem if we have hardlinks, but as they aren't implemented yet and will
probably take a lot work to back-port to HDFS-1.0, we should probably just
ignore that solution.
I'm not convinced that the WAL edit approach that it would be any faster to
take the snapshot. Further, using the WAL actually makes the restore much more
complicated as you need to replay WAL files to get back to the table state - a
big pain for a restoring to a read only table and generally just adding a lot
of lag when you could just hook-up a bunch of immutable files.
> [brainstorm] Timestamp based snapshots in HBase 0.96
> ----------------------------------------------------
>
> Key: HBASE-6180
> URL: https://issues.apache.org/jira/browse/HBASE-6180
> Project: HBase
> Issue Type: Brainstorming
> Reporter: Jesse Yates
> Fix For: 0.96.0
>
>
> Discussion ticket around doing timestamp based snapshots in HBase as an
> extension/follow-on work for HBASE-6055. The implementation in HBASE-6055 (as
> originally defined) is not sufficient for real-time clusters because it
> requires downtime to take the snapshot.
> Time-stamp based snapshots should not require downtime at the cost of
> achieving global consistency.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
