hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Wangda Tan (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-2800) Should print WARN log in both RM/RMAdminCLI side when MemoryRMNodeLabelsManager is enabled
Date Mon, 03 Nov 2014 21:42:34 GMT

    [ https://issues.apache.org/jira/browse/YARN-2800?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14195154#comment-14195154

Wangda Tan commented on YARN-2800:

Hi [~ozawa],
bq. 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?
Not exactly, actually the RM will fail to start, because we have accessible-node-labels in
queues, and when CS initialization, we will check if such labels existed in node labels manager.
Upon mem-based RMNodelabelsManager and RM restart, CS cannot find labels from node labels
manager, so RM will fail to start entirely.

I'm agree about what you mentioned about it may confuse people since admin may configured
it properly in RM side, and it will be annoying every time run such command in client side.
But I think it is still important to let the client know about this. Of course we can add
it in RM web UI, but user may still not check it -- not all user will check cluster metrics
UI :). So I think we can drop logging in RM admin CLI part and change the RMAdmin PB responses
in a separated task, which will return the actual RMNodeLabelsManager being used in RM side.
And we can log the WARN properly.

Do you have any other ideas?


> 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

View raw message