Tim Armstrong has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/14272 )

Change subject: IMPALA-8795 : Enable event polling by default in dockerized 
tests.
......................................................................


Patch Set 4:

(2 comments)

http://gerrit.cloudera.org:8080/#/c/14272/4/tests/util/event_processor_utils.py
File tests/util/event_processor_utils.py:

http://gerrit.cloudera.org:8080/#/c/14272/4/tests/util/event_processor_utils.py@45
PS4, Line 45: is True
nit: "is True" is not needed.


http://gerrit.cloudera.org:8080/#/c/14272/4/tests/util/event_processor_utils.py@69
PS4, Line 69:       time.sleep(total_timeout)
> The flakiness is coming from the logic in the above function. During parall
My understanding was that the above function was checking the notification ID 
from Hive at the time it was called and waiting until the event processors 
catch up. This seems like it should work fine, regardless of concurrency, so 
long as it's called after the notification is generated.

I'm not understanding why that isn't sufficient - is there some delay between 
the hive statement finishing and get_current_notificationEventId() returning 
the event ID from that Hive statement?

Actually I don't understand why the is_hive_insert bit is necessary, either, 
since I would have expected all the notifications to be incremented on the hive 
side after the insert statement finishes.



--
To view, visit http://gerrit.cloudera.org:8080/14272
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I222b64236060b3c4c2d554e2f10e129984ebe972
Gerrit-Change-Number: 14272
Gerrit-PatchSet: 4
Gerrit-Owner: Anurag Mantripragada <[email protected]>
Gerrit-Reviewer: Anurag Mantripragada <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Tim Armstrong <[email protected]>
Gerrit-Comment-Date: Mon, 23 Sep 2019 21:25:15 +0000
Gerrit-HasComments: Yes

Reply via email to