[
https://issues.apache.org/jira/browse/TEZ-2205?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14372045#comment-14372045
]
Hitesh Shah commented on TEZ-2205:
----------------------------------
[~rohini] At the crux of this, there are effectively a couple of
issues/questions in terms of what is the logical behavior and what is the
expectation? The fix for any approach is probably trivial.
Consider the fact that the user configured both YARN and Tez in a conflicting
manner. i.e. Configured YARN to disable timeline but made Tez use Timeline.
Should Tez:
1) error out due to a conflicting configuration i.e YARN timeline disabled but
Tez ATS logger enabled.
2) should Tez try and use Timeline (even though YARN flag is set to false )
and ignore its failures as needed? This should be ok for the most part except
that I think there are some cases in YARN which are not handled cleanly and end
up causing the app to error out. Also, there were some behavioral changes in
YARN-2375 - see below.
3) Should Tez look for the YARN configuration property and silently ignore the
fact that TimelineATSLogger has been configured but it should not be used?
Also, FWIW, earlier ( before YARN-2375 ), even though Tez invoked
Timeline::postEntities, if the YARN flag was set to false, the YARN library
silently dropped the call.
(2) is probably something that YARN needs to address.
As for Tez, we can go with either (1) or (3). (1) is more clear-cut in terms of
making it very clear to the user in terms of how to configure Tez. (3) merely
hides the fact that something is wrongly configured. Also, to clarify, part of
this stems from what is the "yarn.timeline-service.enabled" flag meant to be
used for? Is it a admin flag to control where timeline is enabled or disabled
for the whole cluster? It currently is a client-side flag that cannot be
enforced at all. Furthermore, if it is meant to be used on a per job basis,
should it then be a tez-specific setting ( which we already have in the form of
the class setting ).
Last question for [~rohini]: does the issue of disabling ATS stem from the fact
that it is a bit "hard" to disable ATS logging via the service class name
property?
> Tez still tries to post to ATS when yarn.timeline-service.enabled=false
> -----------------------------------------------------------------------
>
> Key: TEZ-2205
> URL: https://issues.apache.org/jira/browse/TEZ-2205
> Project: Apache Tez
> Issue Type: Sub-task
> Affects Versions: 0.6.1
> Reporter: Chang Li
> Assignee: Chang Li
> Attachments: TEZ-2205.wip.patch
>
>
> when set yarn.timeline-service.enabled=false, Tez still tries posting to ATS,
> but hits error as token is not found. Does not fail the job because of the
> fix to not fail job when there is error posting to ATS. But it should not be
> trying to post to ATS in the first place.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
