You are viewing a plain text version of this content. The canonical link for it is here.
Posted to yarn-issues@hadoop.apache.org by "Sunil G (JIRA)" <ji...@apache.org> on 2017/01/03 06:09:58 UTC
[jira] [Commented] (YARN-6008) Fetch container list for failed
application attempt
[ https://issues.apache.org/jira/browse/YARN-6008?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15794263#comment-15794263 ]
Sunil G commented on YARN-6008:
-------------------------------
Let me take this discussion ahead.
As per current behavior, all running containers of *current attempt* will be returned regardless old/new attemptId is passed via CLI or api. Its makes sense because we are transferring old running containers from killed/failed attempt to new attempt as possible. So it makes sense to show containers of running attempt.
I would like to know whether this makes sense or if its not fine, you could help share specific use case issue where its breaks some of client scenarios. In that case, we can see whether a *cli sub-option* can be provided (for attempt cli/api) to get exclusive containers of given attempt ID (like show an error message that attempt is not running etc). Thoughts? . cc/ [~davidyan]
> Fetch container list for failed application attempt
> ---------------------------------------------------
>
> Key: YARN-6008
> URL: https://issues.apache.org/jira/browse/YARN-6008
> Project: Hadoop YARN
> Issue Type: Bug
> Environment: hadoop version 2.6
> Reporter: Priyanka Gugale
>
> When we run command "yarn container -list" on using failed application attempt we should either get containers from that attempt or get a back list as containers are no longer in running state.
> Steps to reproduce:
> 1. Launch a yarn application.
> 2. Kill app master, it tries to restart application with new attempt id.
> 3. Now run yarn command,
> yarn container -list <Application Attempt ID>
> Where Application Attempt ID is of failed attempt,
> it lists the container from next attempt which is in "RUNNING" state right now.
> Expected behavior:
> It should return list of killed containers from attempt 1 or empty list.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org