Konstantinos Karanasos commented on YARN-2885:

Adding some more on point #2 (I agree with the rest)...
First I agree that the AM should not know whether a container came from the RM 
or from a distributed scheduler.
Regarding the AllocateRequest, I don't think it is currently used in the code, 
so it can be removed.
However, it is used in the RegisterAMRequest to make sure that both the NM and 
the RM have distributed scheduling enabled when setting some of the parameters 
related to the dist scheduling. If we assume that all nodes have dist 
scheduling enabled as long as it is enabled by the RM, then keeping the 
isDistributedScheduling boolean in the RegisterRequest is not needed either. 
After all it is only for setting a few parameters (even if we want to disable 
dist scheduling in a particular NM, that NM can simply discard these 

That said, I am not sure if it is required to create a wrapper at this point 
for the AM protocol.

> Create AMRMProxy request interceptor for distributed scheduling decisions for 
> queueable containers
> --------------------------------------------------------------------------------------------------
>                 Key: YARN-2885
>                 URL: https://issues.apache.org/jira/browse/YARN-2885
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: nodemanager, resourcemanager
>            Reporter: Konstantinos Karanasos
>            Assignee: Arun Suresh
>         Attachments: YARN-2885-yarn-2877.001.patch
> We propose to add a Local ResourceManager (LocalRM) to the NM in order to 
> support distributed scheduling decisions. 
> Architecturally we leverage the RMProxy, introduced in YARN-2884. 
> The LocalRM makes distributed decisions for queuable containers requests. 
> Guaranteed-start requests are still handled by the central RM.

This message was sent by Atlassian JIRA

Reply via email to