[
https://issues.apache.org/jira/browse/TEZ-900?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14301946#comment-14301946
]
chang li commented on TEZ-900:
------------------------------
I have done some further investigation, noticed that the TezSessionManager on
Pig does explicitly call TezAppMasterStatus on TezClient object, which will
print out the diagnostic. Hive currently still have some problems to print out
useful diagnostic on job's failure before launch (such as submit to the wrong
queue), but Hive should be able to solve this easily, because in the "execute"
function from TezTask, monitorExecution will be called, and according to the
@HIVE-5041 diagnostic should be able to printed out. However, in the case of
jobs submitted to the wrong queue, exception will be thrown before
monitorExecution in TezTask's execute function. In the Hive branch-0.13,
exception will be thrown during updateSession, and the waitTillReady within
updateSession. waitTillReady of TezClient will throw an SessionNotRunning
exception, which contains the diagnostic. But currently, Hive TezTask somehow
doesn't print out the exception error to the console, instead record the error
in the log. So the right thing for some Hive people to do is to properly handle
the exception and display the error.
So I propose that every non session job should call TezAppMasterStatus after
TezClient.start, and every session job should call waitTillReady after
TezClient.start. Actually within waitTillReady it will also call the
TezAppMasterStatus. TezAppMasterStatus is an existing function that helps print
out useful diagnostics.
> Confusing message for incorrect queue for some tez examples
> -----------------------------------------------------------
>
> Key: TEZ-900
> URL: https://issues.apache.org/jira/browse/TEZ-900
> Project: Apache Tez
> Issue Type: Bug
> Affects Versions: 0.3.0
> Reporter: Jonathan Eagles
> Assignee: chang li
>
> The error message of at least orderedwordcount with non-existent queue is
> very difficult as compared to mrrsleep which is very intuitive.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
