[
https://issues.apache.org/jira/browse/HIVE-18628?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16625015#comment-16625015
]
Prasanth Jayachandran commented on HIVE-18628:
----------------------------------------------
2 polling interval (200ms and 1000ms) is changed to 1 configurable fixed
interval (500ms). I think with HIVE-16635 the polling interval is 200ms until
dag start. When dag goes to running state, the polling interval is changed to 1
second. Also when there are no WM triggers, there should be no counter
aggregation in AM.
[https://github.com/apache/hive/blob/master/ql/src/java/org/apache/hadoop/hive/ql/exec/tez/monitoring/TezJobMonitor.java#L174]
The making it configurable is intentional change, but making it slow isn't :)
We can certainly change the default to 1000ms to get back the old behaviour
(also not sure if the initial 200ms polling will be useful)
> Make tez dag status check interval configurable
> -----------------------------------------------
>
> Key: HIVE-18628
> URL: https://issues.apache.org/jira/browse/HIVE-18628
> Project: Hive
> Issue Type: Bug
> Affects Versions: 3.0.0
> Reporter: Prasanth Jayachandran
> Assignee: Prasanth Jayachandran
> Priority: Major
> Fix For: 3.0.0
>
> Attachments: HIVE-18628.1.patch
>
>
> DAG status check interval in TezJobMonitor is not configurable. This can
> cause cached data to be returned and fail some trigger validation tests.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
