Wangda Tan commented on YARN-2800:

Moving it to RMStateStore is not a bad idea, since node label store itself can 
be treated as a part of RM's state. However, the RMStateStore is hard coded can 
only use one storage behind, which I'm a little concern about. I'm not quite 
agree with ZK can handle it well, since we shouldn't assume this feature won't 
be used in a large cluster nor high frequency updating labels.

Node labels updating is different from RMStateStore updating, client side can 
change all labels of nodes (like 10k nodes) in one command, but there cannot be 
10k application completed in short period (say around seconds) at least for 
now. WAL based solution may be outperform in such scenario, I think ZK is not a 
good back end for WAL storage.

> Should print WARN log in both RM/RMAdminCLI side when 
> MemoryRMNodeLabelsManager is enabled
> ------------------------------------------------------------------------------------------
>                 Key: YARN-2800
>                 URL: https://issues.apache.org/jira/browse/YARN-2800
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: client, resourcemanager
>            Reporter: Wangda Tan
>            Assignee: Wangda Tan
>         Attachments: YARN-2800-20141102-1.patch, YARN-2800-20141102-2.patch
> Even though we have documented this, but it will be better to explicitly 
> print a message in both RM/RMAdminCLI side to explicitly say that the node 
> label being added will be lost across RM restart.

This message was sent by Atlassian JIRA

Reply via email to