[
https://issues.apache.org/jira/browse/HBASE-4487?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13118201#comment-13118201
]
dhruba borthakur commented on HBASE-4487:
-----------------------------------------
> one cannot guarantee that any individual increment has been committed to the
> WAL.
Even prior to this patch, individual increments were batched-synced-by-hdfs, so
there was no guarantee that each increment operation resulted in a sync.
>From my understanding, the contract for hbase clients is that an increment
>operation is deemed complete only after it has been sycned to the transaction
>log. This contract is valid prior to this patch and this patch does not
>change this contract. The server does not return the RPC response back to the
>client until the increment operation has been sycned to the wal.
Also, the increment operation does not follow the RWCC rules, thereby exposing
uncommitted increment operations to readers. This patch does not change that
behaviour either.
is there something I am missing?
> The increment operation can release the rowlock before sync-ing the Hlog
> ------------------------------------------------------------------------
>
> Key: HBASE-4487
> URL: https://issues.apache.org/jira/browse/HBASE-4487
> Project: HBase
> Issue Type: Improvement
> Reporter: dhruba borthakur
> Assignee: dhruba borthakur
> Fix For: 0.94.0
>
> Attachments: appendNoSync4.txt, appendNoSync5.txt, appendNoSync6.txt
>
>
> This allows for better throughput when there are hot rows.I have seen this
> change make a single row update improve from 400 increments/sec/server to
> 4000 increments/sec/server.
--
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
