[
https://issues.apache.org/jira/browse/MESOS-947?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14071125#comment-14071125
]
Bernd Mathiske commented on MESOS-947:
--------------------------------------
Ben suggested mocking the slave. I have started writing a test in
slave_test.cpp that does just that, and mocks/intercepts runTask(), killTask(),
_runTask(). The general idea is to let runTask() happen, then kill the task
with driver->killTask(), meanwhile let _runTask() do nothing and instead wait
for the kill to complete. Finally I can invoke slave._runTask() with stored
args from the _runTask() invocation that did nothing. This scenario should
eventually work fine. On the way to there, we run into the bug described above.
The killTask() message never actually reaches Slave::killTask(). This is
consistent with the output listed above. Now investigating where exactly we
trap into TASK_LOST, etc.
> Slave should properly handle a killTask() that arrives between runTask() and
> _runTask()
> ---------------------------------------------------------------------------------------
>
> Key: MESOS-947
> URL: https://issues.apache.org/jira/browse/MESOS-947
> Project: Mesos
> Issue Type: Bug
> Reporter: Vinod Kone
> Assignee: Bernd Mathiske
>
--
This message was sent by Atlassian JIRA
(v6.2#6252)
