[
https://issues.apache.org/jira/browse/IMPALA-6956?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16501142#comment-16501142
]
Vuk Ercegovac commented on IMPALA-6956:
---------------------------------------
Looking at the logs, here is a query that expired:
I0531 21:02:26.186826 26134 impala-server.cc:1915] Expiring query
164c0e0c0cffea56:9508677d00000000 due to execution time limit of 1s.
That query corresponds to: SELECT SLEEP(1000000), which was launched here:
I0531 21:02:24.166900 26980 query-state.cc:300] StartFInstances():
query_id=164c0e0c0cffea56:9508677d00000000 #instances=1
At this point, there are two in-flight queries, so this corresponds to the
second query, which was given a 1 second exec_time_limit_s query option.
At the time when that query is expired, the last query was about to be
scheduled. So, it seems that the test just took too long to fire up all queries
prior to the assertion where we expect 4 queries are up and running. Perhaps
the system was overloaded at the time ... will increase that timeout and add
better logging when checking the in-flight queries.
> check_num_executing fails test_query_expiration
> -----------------------------------------------
>
> Key: IMPALA-6956
> URL: https://issues.apache.org/jira/browse/IMPALA-6956
> Project: IMPALA
> Issue Type: Bug
> Components: Backend
> Affects Versions: Impala 3.1.0
> Reporter: Vuk Ercegovac
> Assignee: Vuk Ercegovac
> Priority: Blocker
> Labels: broken-build, flaky
>
> This happened recently on an s3 build. The check that asserts the number of
> currently running queries comes back a few queries short.
> {noformat}
> custom_cluster/test_query_expiration.py:71: in test_query_expiration
> self._check_num_executing(impalad, len(handles))
> custom_cluster/test_query_expiration.py:38: in _check_num_executing assert
> actual == expected
> E assert 2 == 4
> {noformat}
> The last query is executed with:
> {noformat}
> I0426 01:57:29.507251 12229 query-state.cc:395] Executing instance.
> instance_id=be4bd338f1d8a8ef:651e7a3c00000000 fragment_idx=0
> per_fragment_instance_idx=0 coord_state_idx=0 #in-flight=4
> {noformat}
> So there are 4 in flight at that time. However, prior to this, I see that
> some are in the process of being cancelled:
> {noformat}
> I0426 01:57:29.022300 11843 impala-server.cc:1908] Expiring query
> 524c239e0d9f6356:b70c252400000000 due to execution time limit of 1s.
> {noformat}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
