hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Bikas Saha (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-1481) Move internal services logic from AdminService to ResourceManager
Date Sat, 28 Dec 2013 11:13:54 GMT

    [ https://issues.apache.org/jira/browse/YARN-1481?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13858002#comment-13858002
] 

Bikas Saha commented on YARN-1481:
----------------------------------

This method is called by a bunch of refresh* methods() thats are serving user requests. Lets
say the RM is transitioning to standby at this time. Now a refresh*() method gets invoked.
The HA state is checked and found to be active and the method continues to execute and make
changes even though the RM is currently not active.
What is the reason to remove the synchronization?
In general, what I am stressing on is being mindful of race conditions during transitions
of HA state and presenting an error-free view to the users. This may or may not be a problem
here but if we dont approach it in a systematic and consistent manner then we will keep chasing
hard to repro issues in production.

> Move internal services logic from AdminService to ResourceManager
> -----------------------------------------------------------------
>
>                 Key: YARN-1481
>                 URL: https://issues.apache.org/jira/browse/YARN-1481
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>            Reporter: Vinod Kumar Vavilapalli
>            Assignee: Vinod Kumar Vavilapalli
>             Fix For: 2.4.0
>
>         Attachments: YARN-1481-20131207.txt, YARN-1481-20131209.txt, yarn-1481-addendum.patch,
yarn-1481-addendum.patch
>
>
> This is something I found while reviewing YARN-1318, but didn't halt that patch as many
cycles went there already. Some top level issues
>  - Not easy to follow RM's service life cycle
>     -- RM adds only AdminService as its service directly.
>     -- Other services are added to RM when AdminService's init calls RM.activeServices.init()
>  - Overall, AdminService shouldn't encompass all of RM's HA state management. It was
originally supposed to be the implementation of just the RPC server.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)

Mime
View raw message