+1
On 5/10/2016 6:28 AM, Nabarun Nag wrote:
Hi,
I removed the flaky tags from the test in the WAN module. I went through
the commits and made sure they had indeed removed the pauses.
Please do let me know if I have missed something. Or if these these tickets
should still be marked flaky.
ParallelGatewaySenderOperationsDUnitTest:
GEODE-933: Flaky tag was removed - Wait.pause was no longer used.
ParallelWANStatsDUnitTest
GEODE-977: Was Flaky because stats were being read from the wrong VM
Others:
GEODE-1066: Wait criterions were replaced with Awaitility, also order of
creating cache, regions and senders/receivers were modified. No
wait.pauses were used.
GEODE-1147: No wait.pause were present. It was assumed that refactoring of
the WANTestBase and the code had fixed the issue.
GEODE-1148: Flaky tag was removed - Wait.pause was no longer used along
with other modifications to make the test stable.
GEODE-1207: Stopped WAN Locator Discovery Thread when locator is stopped. No
wait.pauses were used.
GEODE-1011: Wait.pauses were removed. Code was modified to stabilize it.
GEODE-1062: Thread.sleep was removed. Issue was resolved by refactoring of
WANTestBase.
GEODE-1032: Listeners were put in to slow down the receivers. No
wait.pauses were used.
GEODE-1121: Memory configurations were changed to stabilize the test. No
wait.pauses were used.
Regards
Nabarun
On Tue, Oct 4, 2016 at 11:04 AM Swapnil Bawaskar <[email protected]>
wrote:
+1
On Tue, Oct 4, 2016 at 10:58 AM, John Blum <[email protected]> wrote:
+1
On Tue, Oct 4, 2016 at 10:11 AM, Kirk Lund <[email protected]> wrote:
Please don't close flaky tickets or remove FlakyTest category unless you
know of a specific commit revision that makes some timing changes to the
test. Unless you replace all the Thread.sleeps with await() calls it's
going to fail again when GC occurs during the test. Just because a test
doesn't fail in 30+ runs, doesn't mean it's not flaky.
ParallelGatewaySenderOperationsDUnitTest has a bunch of these calls:
Wait.pause(2000);
That's pretty much our definition of flaky.
ParallelWANStatsDUnitTest is worse because it has even shorter sleeps:
pause(200);
I'm replacing the sleeps in the ManagementTestBase tests with Awaitility
calls and the tests are dropping from 5 minutes to 30 seconds per dunit
class. So that's another huge motivator to get rid of these broken
sleeps.
-Kirk
On Sun, Oct 2, 2016 at 7:33 PM, Xiaojian Zhou <[email protected]> wrote:
GEODE-933 and GEODE-977 are not reproducible either after run 30+
times.
So
they are not flaky and can be closed for now.
On Sat, Oct 1, 2016 at 11:30 PM, Xiaojian Zhou <[email protected]>
wrote:
1011, 1062, 1066, 1147 have been run 30+ times without reproduce. So
it's
not flaky. I think we can close them. If reproduced someday, we can
re-open.
On Sat, Oct 1, 2016 at 5:09 PM, Anthony Baker <[email protected]>
wrote:
I reviewed a bunch of CI failures today. I closed out duplicates
and
added the ‘CI’ label to JIRA tickets that were missing it. I just
posted a
big review to add the FlakyTest category to bugs with
non-reproducible
failures—pretty much any CI bug that is currently open. Your
comments
are
appreciated (I can push a feature branch if that’s easier):
https://reviews.apache.org/r/52468/
I found several open issues where the flaky category had been
removed.
Can these be marked resolved?
GEODE-933
GEODE-977
GEODE-1011
GEODE-1062
GEODE-1066
GEODE-1147
I have a suspicion that the following open issues are actually
fixed.
Any ideas?
GEODE-1918
GEODE-1333,1334,1335
Anthony
--
-John
503-504-8657
john.blum10101 (skype)
