[
https://issues.apache.org/jira/browse/HBASE-11148?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14044433#comment-14044433
]
Jerry He commented on HBASE-11148:
----------------------------------
Hi, Matteo
Thanks for the comment. I agree.
On the other hand, Procedure is still a good mechanism for centrally kicked
off tasks that need to be executed on RS in parallel.
We can discuss if we need its 'strong consistency' (all have be successful to
report success) -- whether this is needed in many use cases.
We can also improve it to allow and report partial success in the future.
For this particular use case, do you see any potential problem?
Reporting a overall error even when half RS are successful seem a little
'wasteful'. But it is consistent with the name of the API and command.
I can add comments in the API and command let user be aware.
> Provide a distributed procedure to globally roll logs
> -----------------------------------------------------
>
> Key: HBASE-11148
> URL: https://issues.apache.org/jira/browse/HBASE-11148
> Project: HBase
> Issue Type: New Feature
> Reporter: Jerry He
> Assignee: Jerry He
> Fix For: 0.99.0
>
> Attachments: HBASE-11148-trunk-v1.patch, HBASE-11148-trunk-v2.patch
>
>
> Propose a distributed procedure here to globally roll logs.
> Currently HBaseAdmin and HBase shell provides a way to roll the WAL on a
> single RS.
> Some use cases may require that all the RS roll the logs at the same time and
> in a coordinated way. Also there may be requirement that some tasks be done
> together with the roll log on each region server.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
