[
https://issues.apache.org/jira/browse/CASSANDRA-4767?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13535007#comment-13535007
]
Sylvain Lebresne commented on CASSANDRA-4767:
---------------------------------------------
That look good to me overall. The only problem I see is if 2 repairs are
started on the same node simultaneously. In that case, I think both 'nodeProbe'
will get notified and you could get messages for repair you did started which I
suspect would be slightly confusing. Maybe we could make the NodeProbe listener
only listen on the command it has started (i.e. make it discard all other
command numbers). We can't really do that with the repair command since we need
to set up the listener before we call forceTableRepairAsync, but maybe we could
provide a uuid in foceRepairAsync that would be sent back in the notification?
Nit:
- I'm not JMX expert, but the doc for Notification says "It is strongly
recommended that notification senders use the object name rather than a
reference to the MBean object as the source", but as far as I can tell, we do
use the MBean object. No clue why this recommendation though, and everything
seems to work correctly, just mentionning it.
- We're starting to have a bit of duplication of the forceTable* methods in
StorageService. Ideally we could have just one async method that returns a
future, and have the blocking version using it but wait on the future. That
being said, if we just remove the blocking versions in 1.2 it's not a problem
(but I don't know if we want to do that, maybe they are still useful when you
use jconsole for instance).
> Need some indication of node repair success or failure
> ------------------------------------------------------
>
> Key: CASSANDRA-4767
> URL: https://issues.apache.org/jira/browse/CASSANDRA-4767
> Project: Cassandra
> Issue Type: Improvement
> Components: Tools
> Reporter: Ahmed Bashir
> Assignee: Yuki Morishita
> Priority: Minor
> Labels: jmx
> Fix For: 1.1.9
>
> Attachments: 4767-1.1.txt
>
>
> We are currently verifying node repair status via basic log analysis. In
> order to automatically track the status of periodic node repair jobs, it
> would be better to have an indicator (through JMX perhaps).
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
