[
https://issues.apache.org/jira/browse/HBASE-14460?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14905021#comment-14905021
]
Elliott Clark commented on HBASE-14460:
---------------------------------------
bq.Folks did a load of work to avoid our having to wait on sync while the row
lock was held...
Yeah my thought only works after locks are reader/writer.
bq.Don't we still do an mvcc transaction inside the row lock... we'll wait till
mvcc catches up after we take the row lock... then we complete it and on the
end ... So, before releasing the row lock, we'd stamp the mvcc as current
writes sequenceid... unless it had already gone beyond us?
Yeah so that will slow us down in situations where we're highly contented on
the same row. However if there are lots of rows having a check and mutate
operation run on then there is no need to wait of other mvcc transactions to
complete. Currently even if there is no in flight transaction for the row
you're trying to read modify write, the server waits for all transactions to
complete. Holding the lock just gets us up to a point where the server will
only wait on transactions for the row that is in question.
> [Perf Regression] Merge of MVCC and SequenceId (HBASE-HBASE-8763) slowed
> Increments, CheckAndPuts, batch operations
> -------------------------------------------------------------------------------------------------------------------
>
> Key: HBASE-14460
> URL: https://issues.apache.org/jira/browse/HBASE-14460
> Project: HBase
> Issue Type: Bug
> Components: Performance
> Reporter: stack
> Assignee: stack
> Priority: Critical
> Attachments: 14460.txt, region_lock.png
>
>
> As reported by 鈴木俊裕 up on the mailing list -- see "Performance degradation
> between CDH5.3.1(HBase0.98.6) and CDH5.4.5(HBase1.0.0)" -- our unification of
> sequenceid and MVCC slows Increments (and other ops) as the mvcc needs to
> 'catch up' to our current point before we can read the last Increment value
> that we need to update.
> We can say that our Increment is just done wrong, we should just be writing
> Increments and summing on read, but checkAndPut as well as batching
> operations have the same issue. Fix.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
