[
https://issues.apache.org/jira/browse/YARN-1700?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13896150#comment-13896150
]
Zhijie Shen commented on YARN-1700:
-----------------------------------
One arguable issue here is whether we want to record the container that hasn't
been launched. IMO, we should do this even though the recorded information is
not enough to determine the container hasn't be launched yet, as we can enhance
the reported container state and so on to make users understand what happens to
the container better. Another reasons is that we are going to integrate
applicationclientprotocol with applicationhistoryprotocol to make
getcontainer(s) to return active containers as well.
> AHS records non-launched containers
> -----------------------------------
>
> Key: YARN-1700
> URL: https://issues.apache.org/jira/browse/YARN-1700
> Project: Hadoop YARN
> Issue Type: Bug
> Affects Versions: 2.4.0
> Reporter: Gera Shegalov
> Assignee: Gera Shegalov
> Attachments: YARN-1700.v01.patch, YARN-1700.v02.patch
>
>
> When testing AHS with a MR sleep job, AHS sometimes threw NPE out of
> AppAttemptBlock.render because logUrl in container report was null. I
> realized that this is because AHS may record containers that never launch.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
