[
https://issues.apache.org/jira/browse/AURORA-499?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14014181#comment-14014181
]
Bill Farner commented on AURORA-499:
------------------------------------
AFAIK there's nothing we can do save for avoiding aggressive {{git clean}}
invocations. Vagrant stores state in {{.vagrant/}} and otherwise doesn't know
about VMs it previously created.
> Interrupted vagrant provisioning leaves multiple VMs running
> ------------------------------------------------------------
>
> Key: AURORA-499
> URL: https://issues.apache.org/jira/browse/AURORA-499
> Project: Aurora
> Issue Type: Bug
> Components: Testing
> Reporter: Maxim Khutornenko
>
> The following sequence results in a bizarre and non-deterministic behavior
> wrt vagrant testing:
> 1. vagrant up
> 2. ctrl+c before the provisioning finishes
> 3. git clean -fdx and go to step 1 again.
> The above leaves behind multiple VBoxHeadless instances of various readiness.
> All commands, including vagrant ssh, aurora create, scheduler UI refresh and
> etc. result in requests being routed to random boxes. In my case, I was able
> to create a job inside of one VM that routed to the scheduler in another VM.
> The UI was also behaving erratically receiving responses from different boxes
> (mixing old an new UI on different requests).
> This smells like a vagrant bug and I am not sure what exactly we can do here
> short of "pkill -9 VBoxHeadless" before starting a new provisioning.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
