Junping Du commented on YARN-1897:

Thanks for updating the patch, [~mingma]! I take a quick look at the v5 patch 
which looks pretty good to me. However, I got a questions here: I saw we are 
sending SIG via NM-RM heartbeat, and later via container events on NM to pass 
down to each process. Do we think all potential actions on container from user 
can be passed to process in the end? In YARN-4131, I did another more simpler 
way that RM doesn't send signals to NM but send containers preemption/kill 
events directly. Do you think the later way can help in some cases (like tests, 
etc.) or previous way can be enough to handle all cases? i.e. How to extend 
previous way to emulate OOM of container or container LOST/expire, etc.
I will provide more detail comments later. Btw, are failed tests related?

> CLI and core support for signal container functionality
> -------------------------------------------------------
>                 Key: YARN-1897
>                 URL: https://issues.apache.org/jira/browse/YARN-1897
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: api
>            Reporter: Ming Ma
>            Assignee: Ming Ma
>         Attachments: YARN-1897-2.patch, YARN-1897-3.patch, YARN-1897-4.patch, 
> YARN-1897-5.patch, YARN-1897.1.patch
> We need to define SignalContainerRequest and SignalContainerResponse first as 
> they are needed by other sub tasks. SignalContainerRequest should use 
> OS-independent commands and provide a way to application to specify "reason" 
> for diagnosis. SignalContainerResponse might be empty.

This message was sent by Atlassian JIRA

Reply via email to