[
https://issues.apache.org/jira/browse/HBASE-7836?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13634423#comment-13634423
]
Jeffrey Zhong commented on HBASE-7836:
--------------------------------------
That's a good question(my first version was piggyback on the function and
multi). That causes two disadvantages:
1) We need decompose the WAL entries twice one is in the client side to see
where edits should go per region(splitting) and the other decomposing is in the
receiving server side. Inside, replicateWALEntry implementations, all those
WALEntries are also translated to mutations. In addition, inside
replicateWALEntry I'll make the replicateWALEntry a little bit 'messy' to
handle replay case.
2) For better monitoring & priority purpose, a separate command is also better.
The cost to create one command is trivial that's why I choose a new "replay"
command(even with piggyback this command or multi, I still need add special
attributes)
> Create a new "replay" command so that recovered edits won't mess up normal
> coprocessing & metrics
> -------------------------------------------------------------------------------------------------
>
> Key: HBASE-7836
> URL: https://issues.apache.org/jira/browse/HBASE-7836
> Project: HBase
> Issue Type: Sub-task
> Reporter: Jeffrey Zhong
> Assignee: Jeffrey Zhong
> Fix For: 0.95.1
>
> Attachments: hbase-7836-combined.patch, hbase-7836-combined_v2.patch,
> hbase-7836_v1.patch, hbase-7836_v2.patch
>
>
--
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
