[
https://issues.apache.org/jira/browse/ACCUMULO-1049?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13579497#comment-13579497
]
Eric Newton commented on ACCUMULO-1049:
---------------------------------------
The master seems to not notice when tablet servers go away. I have to restart
it.
I'm testing using the continuous ingest test, killing 1-3 servers every 2 5
minutes with 2 minutes between restarts.
> Master does not need to get tserver lock
> ----------------------------------------
>
> Key: ACCUMULO-1049
> URL: https://issues.apache.org/jira/browse/ACCUMULO-1049
> Project: Accumulo
> Issue Type: Improvement
> Components: master
> Affects Versions: 1.4.0
> Reporter: Keith Turner
> Assignee: Keith Turner
> Fix For: 1.5.0, 1.4.3
>
>
> Every tserver has a lock in zookeeper. The master tries to get this lock and
> when it does it assumes the tserver is dead. It seems like this may no
> longer be necessary since changes made in 1.3. The master used to only keep
> track of ip and port to identify a tablet server. It needed to know when it
> got the lock to detect the difference between two tablet server instances at
> the same ip and port.
> Since 1.3 the master keeps distinctive info about a tablet server now and can
> detect the difference between tablet server instances. Making the master
> just monitor the current tablet server locks instead of trying to get the
> lock would simplify the master code.
--
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
