sandflee commented on YARN-4495:

Thanks [~mding],  [~wangda], yes this could simple the code and cover the main 
use case .
1,  we should add containerInfo to InvaildResourceRequest,  should we use a new 
exception like invaildResourceChangeException ? we could just catch this 
2,  I suggest to add a callback like onResourceChangeFailed(List<ContainerId>)  
to make api more user friend.
3,  scheduler may drop request, any suggestion?  check normalizedResource 
before do scheduler.allocate()?
and scheduler may drop container resize request if target resource equals to 
RMContainer allocatedResource, the problem is AM knows nothing about container 
resource normalizition. so: if AM requests resource decrease 8G -> 7.5G, and 
suppose 7.5G is normalized to 8G, rm will drop this request, and will leave AM 
waiting from the reply.

> add a way to tell AM container increase/decrease request is invalid
> -------------------------------------------------------------------
>                 Key: YARN-4495
>                 URL: https://issues.apache.org/jira/browse/YARN-4495
>             Project: Hadoop YARN
>          Issue Type: Improvement
>            Reporter: sandflee
>         Attachments: YARN-4495.01.patch
> now RM may pass InvalidResourceRequestException to AM or just ignore the 
> change request, the former will cause AMRMClientAsync down. and the latter 
> will leave AM waiting for the relay.  

This message was sent by Atlassian JIRA

Reply via email to