Rohith Sharma K S commented on YARN-4205:

bq. Shouldn't it be counted from the time YARN allocates resource for the AM 
and launches it? What if YARN takes more time than the lifetime to allocate 
resource for the app? Seems like the KILL event will be raised immediately 
after the app reaches the RUNNING state in this case. Am I correct?
Basically this point got discussed in earlier comments also, this JIRA is to 
track lifetime of an application. Defining lifetime is nothing but overall 
execution_time of application from submission time. Application get killed at 
any point of time. Basically user is imposing timeout for execution time. Use 
case is every 5 minutes he submits a application and user need output with in 5 
minutes. Use do not worry for time-consumed on statestore or allocation or any 
other which is not user facing also. 
Considering each stages of timeout, in future any other timeouts can be added 
to ApplicationTimeouts class. ApplicationTimeouts java doc also gives 
definition of lifetime.

bq. Should AMRMClientAsync.onShutdownRequest callback be raised to give AM to 
do some last minute work/cleanup/graceful-shutdown-opportunity?
Good point and need to discuss more in general. IIUC, killing an application 
does not allow AM containers to do some clean up. I think the JIRA YARN-2261 is 
intended to do such post clean up.

> Add a service for monitoring application life time out
> ------------------------------------------------------
>                 Key: YARN-4205
>                 URL: https://issues.apache.org/jira/browse/YARN-4205
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: scheduler
>            Reporter: nijel
>            Assignee: Rohith Sharma K S
>         Attachments: 0001-YARN-4205.patch, 0002-YARN-4205.patch, 
> 0003-YARN-4205.patch, 0004-YARN-4205.patch, 0005-YARN-4205.patch, 
> YARN-4205_01.patch, YARN-4205_02.patch, YARN-4205_03.patch
> This JIRA intend to provide a lifetime monitor service. 
> The service will monitor the applications where the life time is configured. 
> If the application is running beyond the lifetime, it will be killed. 
> The lifetime will be considered from the submit time.
> The thread monitoring interval is configurable.

This message was sent by Atlassian JIRA

To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org

Reply via email to