[
https://issues.apache.org/jira/browse/MESOS-4985?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jie Yu updated MESOS-4985:
--------------------------
Labels: mesosphere (was: )
> Destroy a container while it's provisioning can lead to leaked provisioned
> directories.
> ---------------------------------------------------------------------------------------
>
> Key: MESOS-4985
> URL: https://issues.apache.org/jira/browse/MESOS-4985
> Project: Mesos
> Issue Type: Bug
> Affects Versions: 0.28.0
> Reporter: Jie Yu
> Assignee: Gilbert Song
> Priority: Critical
> Labels: mesosphere
> Fix For: 0.28.1
>
>
> Here is the possible sequence of events:
> 1) containerizer->launch
> 2) provisioner->provision is called. it is fetching the image
> 3) executor registration timed out
> 4) containerizer->destroy is called
> 5) container->state is still in PREPARING
> 6) provisioner->destroy is called
> So we can be calling provisioner->destory while provisioner->provision hasn't
> finished yet. provisioner->destroy might just skip since there's no
> information about the container yet, and later, provisioner will prepare the
> root filesystem. This root filesystem will not be destroyed as destroy
> already finishes.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
