[
https://issues.apache.org/jira/browse/AURORA-261?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13935662#comment-13935662
]
Maxim Khutornenko commented on AURORA-261:
------------------------------------------
The initial observation still holds. The scheduler is reacting to the executor
SchedulerMessage by issuing a "delete" action on transition from any terminal
to UNKNOWN state.
> on sandbox garbage collection, just de-link the slave HREF on the finished
> tasks page
> -------------------------------------------------------------------------------------
>
> Key: AURORA-261
> URL: https://issues.apache.org/jira/browse/AURORA-261
> Project: Aurora
> Issue Type: Story
> Components: Scheduler, UI, Usability
> Reporter: brian wickman
> Priority: Minor
>
> Instead of removing the task from the finished tasks log when we get a signal
> from the GC executor, just grey-out the URL to the thermos sandbox in the UI
> instead of removing it from the scheduler database entirely.
> If you have a 12-hourly cron job and you look at Finished Tasks and see:
> 0 hours ago
> 12 hours ago
> 36 hours ago
> 48 hours ago
> 60 hours ago
> 72 hours ago
> It'd be totally reasonable to come to the conclusion that the scheduler had a
> failover at the 24 hour mark, but instead it could just be because the
> machine that happened to get the task 24 hours ago was under strong space
> contention and just got garbage collected much sooner than everyone else.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
