[jira] [Commented] (TEZ-2216) Expose errors during AM initialization

["Hitesh Shah (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/TEZ-2216?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14389498#comment-14389498 ] 

Hitesh Shah commented on TEZ-2216:
----------------------------------

[~lichangleo] The problem stmt is as follows: 

The DAGAppMaster class can be thought of as a composite service consisting of multiple services. The YarnSchedulerService is one of them. Today, if any service fails to init or start, the AM fails without fully unregistering from the RM. 

The objective here is to first pin point the error i.e. which service failed to come up and why? Then, actually start the yarn scheduler service if it has not been started and eventually use it to un-register with the error set to failed and the diagnostics pointing to the error traced earlier ( which service failed to init/start ).

 





> Expose errors during AM initialization
> --------------------------------------
>
>                 Key: TEZ-2216
>                 URL: https://issues.apache.org/jira/browse/TEZ-2216
>             Project: Apache Tez
>          Issue Type: Bug
>            Reporter: Bikas Saha
>         Attachments: TEZ-2216.1.patch
>
>
> If there are bad configs or other issues that cause errors/exceptions during AM initialization (eg. during service init) then those errors are not exposed to the user. Exposing them would be useful in quickly debugging such issues.



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