[
https://issues.apache.org/jira/browse/HBASE-12988?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14609311#comment-14609311
]
Lars Hofhansl commented on HBASE-12988:
---------------------------------------
We are using the writeTime on the sink to calculate ageOfLastAppliedOp, which
is an important metric, I suppose we can ship the max time we've seen in a
batch and still be correct. In that case we can also optimize the replication
stream and only send a WALKey once followed by all WALEdits belong to it.
But... I assume at some point we'll want to make replication more "consistent",
ordering by seqNum (or delaying visibility by seqNum) would be a (the only?)
way to do that. I do not want to design anything that prevents us from doing
this going forward. Grouping by HRegion is simple (this patch is all that is
needed) and (presumably) effective.
> [Replication]Parallel apply edits on row-level
> ----------------------------------------------
>
> Key: HBASE-12988
> URL: https://issues.apache.org/jira/browse/HBASE-12988
> Project: HBase
> Issue Type: Improvement
> Components: Replication
> Reporter: hongyu bi
> Assignee: Lars Hofhansl
> Attachments: 12988-v2.txt, 12988-v3.txt, 12988.txt,
> HBASE-12988-0.98.patch, ParallelReplication-v2.txt
>
>
> we can apply edits to slave cluster in parallel on table-level to speed up
> replication .
> update : per conversation blow , it's better to apply edits on row-level in
> parallel
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
