[
https://issues.apache.org/jira/browse/SLIDER-341?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14106271#comment-14106271
]
Sumit Mohanty commented on SLIDER-341:
--------------------------------------
That would be ideal. Anyway here is the scenario and I am wondering how to
merge it with the container failure window/threshold support.
Storm components are set to auto-restart. These are auto-restart within the
same container (there is a need to preserve local resources as much as
possible).
We could count each re-start as a container failure by having the
*ProviderService report it as a failure to AppState.
One requirement could be that if a user wants to try to come up on the same
host/container for X times before re-allocating the container itself.
Essentially, what we are saying is that auto-restart on the same container
means that the component does not want to move once started and is OK to have
the application stopped if it fails too many times. The user may still force a
move by flexing down and then flexing up (while getting rid of the host).
> Add a window based failure count for auto-start to limit indefinite attempt
> ---------------------------------------------------------------------------
>
> Key: SLIDER-341
> URL: https://issues.apache.org/jira/browse/SLIDER-341
> Project: Slider
> Issue Type: Bug
> Components: agent, app-package
> Affects Versions: Slider 0.50
> Reporter: Sumit Mohanty
> Assignee: Sumit Mohanty
>
> The agent's auto-restart feature for component instance should be gated by a
> window based upper limit.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
