[
https://issues.apache.org/jira/browse/SLIDER-403?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14124535#comment-14124535
]
Steve Loughran commented on SLIDER-403:
---------------------------------------
I don't think we need to do this ... you can just edit the threshold in
resources.json during development
{code}
"yarn.container.failure.threshold":"1",
"yarn.container.failure.window.hours":"1"
{code}
see {{content/docs/slider_specs/resource_specification.md}}
> Handle container failure threshold dynamically
> ----------------------------------------------
>
> Key: SLIDER-403
> URL: https://issues.apache.org/jira/browse/SLIDER-403
> Project: Slider
> Issue Type: Improvement
> Reporter: Ted Yu
>
> Basic windowed failure handling is in place.
> Steve made the following suggestion:
> You could specify a smaller number for failure threshold and still
> distinguish "broken app/configuration" from "unreliable cluster".
> During app package development, a failure threshold of 1 is probably
> appropriate
> basic windowed = we reset the counter after a configured interval. Whenever
> the cluster is flexed the counter is also reset
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
