[
https://issues.apache.org/jira/browse/MESOS-351?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13578849#comment-13578849
]
Vinod Kone commented on MESOS-351:
----------------------------------
Also, I think this is only half of the story. Because, even with detector doing
the correct thing, the master would've simply re-registered this slave and let
it continue. This should be solved with the "registrar" that stores some
persistent state about the master.
> Master detector should not block on zookeeper 'get' method
> ----------------------------------------------------------
>
> Key: MESOS-351
> URL: https://issues.apache.org/jira/browse/MESOS-351
> Project: Mesos
> Issue Type: Bug
> Reporter: Vinod Kone
>
> Currently, the master detector code calls zk->get() in its detectMaster()
> method. Under the covers, we call zoo_aget(...., callback,....) and block
> till the callback is called.
> At Twitter, we encountered a situation where zk->get() got blocked because
> the callback was never called by the underlying zookeeper library. Even if a
> watch expired indicating session expired, the detector would not process it
> because it is blocked in detectMaster().
> I will add some data below.
--
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
