[jira] [Resolved] (IMPALA-6242) Flaky test: TimerCounterTest.CountersTestOneThread

2018-02-02 Thread Tianyi Wang (JIRA)

 [ 
https://issues.apache.org/jira/browse/IMPALA-6242?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Tianyi Wang resolved IMPALA-6242.
-
Resolution: Fixed

IMPALA-6242: Change runtime-profile-test into using the same clock

In runtime-profile-test, both MonotonicStopWatch::Now() and
MonotonicNanos() are used. The former may use CLOCK_MONOTONIC_COARSE or
CLOCK_MONOTONIC while the latter always uses CLOCK_MONOTONIC. This may
contribute to the flakiness of this test. This patch changes this test
into using MonotonicStopWatch::Now() uniformly.

Change-Id: I20648d88e666b08d6f2a67a4158244865a50b629
Reviewed-on: http://gerrit.cloudera.org:8080/9171
Reviewed-by: Alex Behm 
Tested-by: Impala Public Jenkins

> Flaky test: TimerCounterTest.CountersTestOneThread
> --
>
> Key: IMPALA-6242
> URL: https://issues.apache.org/jira/browse/IMPALA-6242
> Project: IMPALA
>  Issue Type: Bug
>  Components: Backend
>Affects Versions: Impala 2.11.0, Impala 2.12.0
>Reporter: Tianyi Wang
>Assignee: Tianyi Wang
>Priority: Critical
>  Labels: broken-build
> Fix For: Impala 2.12.0
>
>
> CountersTestOneThread failed in an release build. The difference between the 
> time interval measured in spawned thread and that in main thread is more than 
> expected.
> {noformat}
> be/src/util/runtime-profile-test.cc:673
> Expected: (stopwatch_value) >= (expected_value - 
> TimerCounterTest::MAX_TIMER_ERROR_NS), actual: 250997548 vs 258086034
> {noformat}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)


[jira] [Resolved] (IMPALA-6242) Flaky test: TimerCounterTest.CountersTestOneThread

2017-12-12 Thread Tianyi Wang (JIRA)

 [ 
https://issues.apache.org/jira/browse/IMPALA-6242?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Tianyi Wang resolved IMPALA-6242.
-
Resolution: Fixed

IMPALA-6242: Reduce flakiness in TimerCounterTest

The error threshold in TimerCounterTest is 15ms, which is still not
enough in some rare cases. This patch increases it to 30ms.

Change-Id: Ifc038908857060ccbabfe30c46e72fd93907f412
Reviewed-on: http://gerrit.cloudera.org:8080/8670
Reviewed-by: Tim Armstrong 
Tested-by: Impala Public Jenkins


> Flaky test: TimerCounterTest.CountersTestOneThread
> --
>
> Key: IMPALA-6242
> URL: https://issues.apache.org/jira/browse/IMPALA-6242
> Project: IMPALA
>  Issue Type: Bug
>  Components: Backend
>Affects Versions: Impala 2.11.0
>Reporter: Tianyi Wang
>Assignee: Tianyi Wang
>Priority: Critical
>  Labels: broken-build
>
> CountersTestOneThread failed in an release build. The difference between the 
> time interval measured in spawned thread and that in main thread is more than 
> expected.
> {noformat}
> be/src/util/runtime-profile-test.cc:673
> Expected: (stopwatch_value) >= (expected_value - 
> TimerCounterTest::MAX_TIMER_ERROR_NS), actual: 250997548 vs 258086034
> {noformat}



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)