[
https://issues.apache.org/jira/browse/AURORA-86?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13880556#comment-13880556
]
Jeff Currier commented on AURORA-86:
------------------------------------
That's correct. This work has already begun (in the process of
implementation). The expectation is that the automated self-healing should
help reduce the overall impact of any one machine death by executing a series
of automated repairs. In the cases where these do not finally, "fix" the issue
then the machine will be isolated out of use by any framework.
> Explore mitigation strategies for "machines of death"
> -----------------------------------------------------
>
> Key: AURORA-86
> URL: https://issues.apache.org/jira/browse/AURORA-86
> Project: Aurora
> Issue Type: Task
> Components: Scheduler
> Reporter: Bill Farner
> Priority: Minor
>
> Hung or otherwise bad machines occasionally crop up. It's worth doing the
> thought exercise to consider ways we might detect it ourselves so tasks are
> not black-holed indefinitely or sent to a machine that they are likely to
> fail on.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
