[
https://issues.apache.org/jira/browse/TEZ-4171?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17102311#comment-17102311
]
Gopal Vijayaraghavan commented on TEZ-4171:
-------------------------------------------
bq. If the client and server states are different, the server can send a
response immediately, otherwise we sleep until signaled or timeout occurs.
This is what is currently the problem but only for one intermediate state, as
the "response immediately" is only done for terminal states of the query - the
transition from SUBMITTED to RUNNING is being flipped to "respond immediately",
at least for the next rpc call.
The only definite information we know is that we've yet to tell the client
about the RUNNING state, which is the assumption this patch is going against.
> DAGImp::getDAGStatus should try to report RUNNING state information correctly
> -----------------------------------------------------------------------------
>
> Key: TEZ-4171
> URL: https://issues.apache.org/jira/browse/TEZ-4171
> Project: Apache Tez
> Issue Type: Improvement
> Reporter: Rajesh Balamohan
> Assignee: Rajesh Balamohan
> Priority: Major
> Attachments: TEZ-4171.1.patch, TEZ-4171.2.patch, TEZ-4171.3.patch,
> TEZ-4171.4.patch, TEZ-4171.5.patch, TEZ-4171.6.patch
>
>
> [https://github.com/apache/tez/blob/master/tez-dag/src/main/java/org/apache/tez/dag/app/dag/impl/DAGImpl.java#L593]
>
> [https://github.com/apache/tez/blob/master/tez-dag/src/main/java/org/apache/tez/dag/app/dag/impl/DAGImpl.java#L953]
>
> It is possible that the state changed to "RUNNING" and signaling was done.
> However, if getDAGStatus gets invoked later, it wouldn't receive any signal
> and ends up waiting until the timeout. For clients trying to report details
> on SUBMIT–>RUNNING, it ends up giving wrong picture.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
