[
https://issues.apache.org/jira/browse/MESOS-2987?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14612095#comment-14612095
]
Alexander Rukletsov commented on MESOS-2987:
--------------------------------------------
IMO, in this case, a slave should fail before it even tries to recover
frameworks and executors. We already have a failure with the message
"Checkpointed resources are incompatible with slave resources" and I would say
we should hit this path and fail with this message instead of current behaviour.
> Fail earlier on startup when starting mesos slave after changing resources
> --------------------------------------------------------------------------
>
> Key: MESOS-2987
> URL: https://issues.apache.org/jira/browse/MESOS-2987
> Project: Mesos
> Issue Type: Bug
> Affects Versions: 0.22.1
> Environment: CentOS 7
> Reporter: Craig W
> Priority: Minor
>
> If the slave is running, then I modify the resources and restart the slave I
> see "Failed to load executore forked pid file..." in the logs and the slave
> fails to start.
> If instead the slave is running, then I modify the resources, and I remove
> meta/slaves/latest/slave.info, then I can restart the slave successfully.
> The mesos-slave should fail earlier in this case and provide a better message
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
