Wangda Tan commented on YARN-3717:

Thanks for thinking about this, mostly makes sense to me. Now we lack of some 
necessary information showing on web page that makes node label debugging is 

For RM logs, I think most of them are available in debug mode. Such as 
user-limit, am-resource-limit, queue-limit, etc. It may be hard to add them to 
log level, because if a job is hung, every node heartbeat cannot allocate 
resource for the job. If the cluster has lots of nodes, the log message # will 
be huge.

I suggest to focus on node label RM web UI changes in this ticket. I will edit 
title/desc a little bit to make sure it reflects what we want to do.

> Improve debug-ability of Node Labels
> ------------------------------------
>                 Key: YARN-3717
>                 URL: https://issues.apache.org/jira/browse/YARN-3717
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>            Reporter: Naganarasimha G R
>            Assignee: Naganarasimha G R
>         Attachments: RMLogsForHungJob.log
> Few improvements which i want to suggest are : 
> 1> Add the default-node-Label expression for each queue in WebUI 
> 2> In Application/Appattempt page  show the app configured node label 
> expression for AM and Job (may be we need to store in ATS too)
> 3> RM log for the application ID which has hung has been attached, it is 
> difficult to analyze as there is no information related to the labels in it . 
> I was intending to add @
> "2015-05-26 00:19:25,681 INFO 
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.LeafQueue: 
> Application added - appId: application_1432579630889_0001 user: 
> org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.LeafQueue$User@35b4193e,
>  leaf-queue: default #user-pending-applications: 0 #user-active-applications: 
> 1 #queue-pending-applications: 0 #queue-active-applications: 1" open for 
> suggestions for adding Info logs

This message was sent by Atlassian JIRA

Reply via email to