[ 
https://issues.apache.org/jira/browse/YARN-4451?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15055837#comment-15055837
 ] 

Naganarasimha G R commented on YARN-4451:
-----------------------------------------

bq. Can you do a sizing test to see how big the log is?
It took around 3MB for 5 mins in 7 node cluster with one app, (and for 3 node 
cluster also more or less the same with single app)

bq. What http method are you using - that url only supports POST.
My Mistake. i tried with get . Also if time is not passed do we need to 
consider default as 1 min ? 

How abt other question :
Also we observed that DEBUG logs were only present and not the INFO logs and 
usually we do not put debug logs for the same info log message, so was 
wondering whether its feasible to collect INFO logs too in the same log file so 
that analysis is faster ?

> Some improvements required in Dump scheduler logs
> -------------------------------------------------
>
>                 Key: YARN-4451
>                 URL: https://issues.apache.org/jira/browse/YARN-4451
>             Project: Hadoop YARN
>          Issue Type: Improvement
>            Reporter: Naganarasimha G R
>            Assignee: Naganarasimha G R
>
> Though dumping scheduler logs is very useful option there are few nits in 
> using it
> * for naive or first time user its hard to understand what does {{"Time"}} 
> stand for past or future, IMO it would be slightly better to set the name in 
> the ui as {{"Time Period"}}
> * success message should give where the logs will be found and the file name
> * Need to append the time stamp and the period to the file name, so that its 
> not over ridden
> * From code it seems like it always returns {{"Capacity scheduler logs are 
> being created"}} even though the fair scheduler is set
> * Would having cli option in {{"yarn rmadmin"}}  will also be helpful ? 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to