I agree that we are ready for the RC, and that deflaking the tests is great.
However, I think that Permissions test failures are exposing a real problem (probably in HBase), and we've only adopted the tests for the buggy behaviour of HBase, and should track down / fix the root cause. On Sun, May 30, 2021 at 4:46 PM Viraj Jasani <[email protected]> wrote: > > IMHO, a flapper is worse than a failing test because it undermines the > confidence in a test run. > > I concur. In fact, in the last few builds, I have seen a few flappers (e.g > testUpsertSelectWithMultiByteCharsAutoCommit, testAsyncRebuildAll) whose > resolution was attempted just a few months back and now they are showing up > again (with much lower frequency than before), so yes we need some > attention again. > > After fixing consistent failure with testPherfMain (PHOENIX-6482) and high > frequency flappers with BasePermissionsIT and AuditLoggingIT > (PHOENIX-6483), CI builds are in better shape now. > In the recent build, I can see no test failures with 3/4 HBase profiles. I > will file a Jira for JaCoCo report generation failure, which keeps showing > up often (flapper with JaCoCo issue, not related to test failure). > > Given that CI builds are in better shape, planning to start an RC soon. > Thanks > > > On Sat, May 29, 2021 at 9:45 PM [email protected] <[email protected]> wrote: > > > Nice! > > > > I noticed we had some successful 5.1 runs too. > > Fixing the tests would need some attention again. Flappers should either > > be fixed or disabled. > > > > IMHO, a flapper is worse than a failing test because it undermines the > > confidence in a test run. > > If the tests always pass you'll notice a failure right away. > > But when most of the runs produce some kind of failure it is human nature > > to start ignoring them, and then new, real failures pile up. > > > > > > -- Lars > > > > > > On Friday, May 28, 2021, 11:36:53 PM PDT, Viraj Jasani < > [email protected]> > > wrote: > > > > > > > > > > > > Due to incorrect/missing fix versions, found some discrepancies b/ master > > and 5.1 in the Pherf module. > > Backported recent improvements to 5.1 including test fix: > > PHOENIX-6417, PHOENIX-6118, PHOENIX-6430, PHOENIX-6429, PHOENIX-6431, > > PHOENIX-6432 > > > > On Sat, May 29, 2021 at 12:20 AM Viraj Jasani <[email protected]> > wrote: > > > > > I looked at CI build today and there are a couple of flakies in > > core/pherf. > > > > > > 3 major flaky tests: PermissionNSEnabledIT, PermissionsCacheIT, > > > AuditLoggingIT (reported with HBase 2.2/2.3/2.4). Permission tests have > > > been flaky for quite some time. IIRC, there was some attempt to fix > them > > as > > > well but still they appear with AccessDeniedException at times. Not > much > > > sure about AuditLoggingIT though. > > > > > > Pherf test fails with HBase 2.1. > > > > > > > > > On Fri, 28 May 2021 at 10:44 PM, [email protected] <[email protected]> > > > wrote: > > > > > >> 5.1 CI builds seems to fail consistently - at least for the past two > > >> weeks. > > >> > > >> > > > https://ci-hadoop.apache.org/job/Phoenix/job/Phoenix-mulitbranch/job/5.1/ > > >> > > >> Might be something to look into. > > >> I have an hour today between meetings, I'll see what I can do. :) > > >> > > >> -- Lars > > >> > > >> > > >> On Wednesday, May 26, 2021, 10:54:13 PM PDT, Viraj Jasani < > > >> [email protected]> wrote: > > >> > > >> > > >> > > >> > > >> > > >> Thanks Lars. I had an offline chat with Istvan yesterday and I will > > >> prepare > > >> RCs over the weekend. > > >> > > >> In the meantime, I will again take a look at Jira fixVersion and git > > >> commits compatibility and also try to help with any backport (if still > > >> pending). > > >> > > >> > > >> On Wed, 26 May 2021 at 11:53 PM, [email protected] <[email protected]> > > >> wrote: > > >> > > >> > I cherry-picked three changes into branch 5.1. > > >> > Assuming Jira is up-to-date, there are no 4.16.x issues that are not > > >> also > > >> > in 5.1.x. > > >> > > > >> > There are 12 issues left that are in 4.17.0 and 5.2.0, but not in > > 5.1.x. > > >> > But they are presumably not in 4.16.x because they violate backward > > >> > compatibility, and - if so - should not be in 5.1.2. > > >> > > > >> > -- Lars > > >> > > > >> > On Wednesday, May 26, 2021, 10:15:31 AM PDT, [email protected] < > > >> > [email protected]> wrote: > > >> > > > >> > > > >> > > > >> > > > >> > > > >> > Awesome. Thanks Istvan. > > >> > > > >> > I'll do a pass through the issues too. > > >> > > > >> > -- Lars > > >> > > > >> > > > >> > On Tuesday, May 25, 2021, 8:45:00 PM PDT, Istvan Toth < > > [email protected] > > >> > > > >> > wrote: > > >> > > > >> > > > >> > > > >> > > > >> > > > >> > I am not aware of any blockers in 5.1. > > >> > If there is no objection, and no other volunteers for the 5.1.2 RM > > >> role, I > > >> > intend to start the process on Monday. > > >> > > > >> > I will also look for missing fixes before release, but if anyone is > > >> aware > > >> > of any fix that is needed in 5.1.2 , but > > >> > hasn't been backported, then please try to get it resolved by > Monday. > > >> > > > >> > regards > > >> > Istvan > > >> > > > >> > On Wed, May 26, 2021 at 4:19 AM [email protected] <[email protected]> > > >> wrote: > > >> > > > >> > > There are 17 resolved issues against it. > > >> > > PHOENIX-6436 is needed for further integration with Trino > (formerly > > >> > > Presto). > > >> > > > > >> > > It is time for another point release? > > >> > > > > >> > > -- Lars > > >> > > > > >> > > > >> > > > > > >
