You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@myriad.apache.org by "Santosh Marella (JIRA)" <ji...@apache.org> on 2015/10/08 22:33:26 UTC

[jira] [Commented] (MYRIAD-124) NM launched not connecting to RM

    [ https://issues.apache.org/jira/browse/MYRIAD-124?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14949339#comment-14949339 ] 

Santosh Marella commented on MYRIAD-124:
----------------------------------------

Will be fixed as part of NM Placement (MYRIAD-105)

> NM launched not connecting to RM
> --------------------------------
>
>                 Key: MYRIAD-124
>                 URL: https://issues.apache.org/jira/browse/MYRIAD-124
>             Project: Myriad
>          Issue Type: Bug
>          Components: Executor, Scheduler
>         Environment: All
>            Reporter: Aashreya Ravi Shankar
>            Assignee: Santosh Marella
>
> I was trying out the latest commit of FGS, here is what I am facing.
> When I launch the RM, I need to specify the -Dyarn.resourcemanager.hostname=<hostnameOfRM>
> When I did not specify this during RM launch, and try to Flex up. NodeManagers may get launched on nodes not running the RM. In this case, the NM does not recognize the RM, it tries to look for RM running on the same machine. Once such an NM gets launched I am unable to Flex it down, gets hung in the same state and end up manually killing it.
> We need a fall-back mechanism to identify the RM incase the hostname is not specified, to avoid this situation.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)