Tsuyoshi OZAWA commented on YARN-2800:

[~leftnoteasy], thanks for your comments.

+              + "this message is based on the yarn-site.xml settings "
+              + "in the machine you run \"yarn rmadmin ...\", if you "
+              + "already edited the field in yarn-site.xml of the node "
+              + "running RM, please ignore this message.";

I think printing the message based on client-side configuration can confuse the 
user - it can be different from RM-side. Every users doesn't have a copy of 
RM-side configuration and some users doesn't know the contents of RM-side 

But if user configured mem-based node labels manager, user may add labels to 
queue configurations, when RM will be failed to launch (specifically, CS cannot 
initialize) if a queue use a label but not existed in node labels manager

Let me clarify this case - do you mean RM will fail to allocate containers on 
labeled nodes after RM restart since RM uses MemoryRMNodeLabelsManager and 
forget the mapping of node-to-labels? In this case, I think we should arise the 
warnings to submitter of yarn apps like "application cannot be submitted for 
now since no node has the required label" after restart. It's more straight 
forward because users can notice the mistake of configurations of labels.

So I think it's better way to log the warning at startup once and add the 
information to Web UI for the consistency of the information. What do you think?

> 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