[
https://issues.apache.org/jira/browse/HBASE-12751?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14378131#comment-14378131
]
stack commented on HBASE-12751:
-------------------------------
Excellent, excellent stuff (this whole issue and this finding)
So, first response is same as yours: Add to memstore after WAL; the edits can't
be 'seen' yet anyways because they are beyond the read point.... But our
current ordering of memstore first was about perf so would need to look at this.
Other things are going on here though FYI, our [~busbey] is looking at WAL perf
in the layer below DFS client and is having some radical thoughts... so this is
right time to raise reorder on the append pipeline; longer time between syncs
means more grouping... higher throughput.
You looking at the rerdering then [~eclark]?
> Allow RowLock to be reader writer
> ---------------------------------
>
> Key: HBASE-12751
> URL: https://issues.apache.org/jira/browse/HBASE-12751
> Project: HBase
> Issue Type: Bug
> Components: regionserver
> Reporter: Elliott Clark
> Assignee: Nate Edel
> Attachments: HBASE-12751-v1.patch, HBASE-12751.patch
>
>
> Right now every write operation grabs a row lock. This is to prevent values
> from changing during a read modify write operation (increment or check and
> put). However it limits parallelism in several different scenarios.
> If there are several puts to the same row but different columns or stores
> then this is very limiting.
> If there are puts to the same column then mvcc number should ensure a
> consistent ordering. So locking is not needed.
> However locking for check and put or increment is still needed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
