[
https://issues.apache.org/jira/browse/TEZ-3128?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Siddharth Seth updated TEZ-3128:
--------------------------------
Attachment: amJstack
Here's the stack trace. TezContainerLauncher invokes shutdownAllContainers -
which seems to try and kill the containers - one at a time - before shutdown.
Releasing containers is sufficient. [~hitesh] - last I spoke to you about this,
I got the impression that a stop container was required for some reason. If
that understanding was incorrect, we should be able to avoid the container
stop. Do we even need to release containers ? If the app is shutting down, YARN
should take care of this on it's own - once the application unregisters ? (or
is that where the problem is - the unregistration happens at the end and hence
we should release the containers early).
IIRC, even for unregistration, YARN allows an app to unregister early - and
will not kill the AM for a certain amount of time after that.
> Avoid stopping containers on the AM shutdown thread
> ---------------------------------------------------
>
> Key: TEZ-3128
> URL: https://issues.apache.org/jira/browse/TEZ-3128
> Project: Apache Tez
> Issue Type: Bug
> Affects Versions: 0.8.0-alpha
> Reporter: Siddharth Seth
> Assignee: Tsuyoshi Ozawa
> Labels: newbie
> Attachments: TEZ-3128.001.patch, amJstack
>
>
> During an AM shutdown, the TaskCommunicator is also shutdown and it tries to
> stop containers in the shutdown thread itself. This can cause the AM shutdown
> to block if NMs are not available.
> This likely affects 0.7 as well.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
