Hi Chris, I believe we’ve finally addressed the flaky unit tests you were seeing in #8237.
The Yetus run came back completely green Already fixed unit tests (merged): - hadoop.yarn.sls.appmaster.TestAMSimulator - hadoop.yarn.server.router.subcluster.fair.TestYarnFederationWithFairScheduler - hadoop.yarn.server.router.webapp.TestFederationWebApp - hadoop.yarn.server.router.webapp.TestRouterWebServicesREST - hadoop.yarn.server.resourcemanager.webapp.TestRMWebServicesReservation Additional flaky tests that have been fixed and are now awaiting review: - hadoop.hdfs.tools.TestDFSAdmin → https://github.com/apache/hadoop/pull/8269 - hadoop.yarn.server.resourcemanager.TestRMHA → https://github.com/apache/hadoop/pull/8267 - hadoop.yarn.service.TestYarnNativeServices → https://github.com/apache/hadoop/pull/8266 - org.apache.hadoop.mapreduce.v2.TestUberAM → https://github.com/apache/hadoop/pull/8263 Would you mind taking a look at the open PRs when you have a moment? Thanks a lot! Best regards, Shilun Fan. On Wed, Feb 18, 2026 at 1:52 PM Chris Nauroth <[email protected]> wrote: > I'll be away the rest of this week, but I'm planning to create a 3.5.0 RC > as soon as I get back the week of 2/23. > > I'm also nearly done verifying the new 3.4.3 RC, just waiting on a few more > tests. > > Chris Nauroth > > > On Mon, Feb 16, 2026 at 4:14 PM Chris Nauroth <[email protected]> wrote: > > > At this point, I think these are just extremely flaky tests. I've tried > > numerous git bisect exercises to pinpoint specific commits. No matter > > what I do though, I never see a consistent pass or a consistent fail. > > > > Patches are welcome to stabilize the tests, but I won't treat these as > > release 3.5.0 blockers unless I hear otherwise. > > > > Chris Nauroth > > > > > > On Fri, Feb 13, 2026 at 3:05 PM Chris Nauroth <[email protected]> > wrote: > > > >> So far I haven't been able to connect these test failures to any > specific > >> commits. I reverted my local copy all the way back to July, and the > tests > >> still failed. Maybe this is more like some ticking time bomb that's been > >> present in the code for a long time rather than a recently introduced > bug. > >> > >> YARN-11926 reports some bad test data (old timestamps). That might > >> partially explain it. > >> > >> Chris Nauroth > >> > >> > >> On Fri, Feb 13, 2026 at 10:18 AM Steve Loughran <[email protected]> > >> wrote: > >> > >>> Thanks everyone. I've got the 3.4.3 RC1 done but going to play with it > >>> myself over the weekend. > >>> > >>> I made the mistake of trying to get google gemini cli to write a test > in > >>> a > >>> two class project while doing the build and now need to lie down > rather > >>> than look at an IDE > >>> > >>> " My apologies for neglecting GEMINI.md guidelines. I must revert > >>> System.out.println and SLF4J logging, and remove reflection-based > >>> injection > >>> from TestCatalogSigner.java. My focus will now be on understanding why > >>> S3V4RestSignerClient.create(props) returns null without Mockito or > >>> reflection, potentially rethinking the test approach if a non-null > >>> instance > >>> is impossible without a live service. Starting with restoring SLF4J > >>> logging in CatalogSigner.java." > >>> > >>> > >>> > >>> On Fri, 13 Feb 2026 at 10:37, Xiaoqiao He <[email protected]> > wrote: > >>> > >>> > Thank you both for the great work. About test failure #TestDFSAdmin, > it > >>> > looks that > >>> > this thread[1] does not finish as expected, but I did not dig where > >>> code > >>> > changes > >>> > trigger this failure now. It should be fixed or marked before > release. > >>> > Thanks again. > >>> > > >>> > [1] > >>> > > >>> > > >>> > https://github.com/apache/hadoop/blob/trunk/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/tools/TestDFSAdmin.java#L1254 > >>> > > >>> > Best Regards, > >>> > - He Xiaoqiao > >>> > > >>> > On Fri, Feb 13, 2026 at 1:42 PM Chris Nauroth <[email protected]> > >>> wrote: > >>> > > >>> > > Awesome, I appreciate your help! I'll keep investigating the > >>> remaining > >>> > > issues. > >>> > > > >>> > > Chris Nauroth > >>> > > > >>> > > > >>> > > On Thu, Feb 12, 2026 at 4:26 PM slfan1989 <[email protected]> > >>> wrote: > >>> > > > >>> > > > Hi Chris, > >>> > > > > >>> > > > Thanks for driving the Hadoop 3.5.0 release forward. > >>> > > > > >>> > > > +1 from me on HADOOP-19811 / PR #8243. > >>> > > > > >>> > > > I’ll also take a look at the current trunk test failures, > >>> especially > >>> > the > >>> > > > YARN-related unit test failures, and report back with findings > or a > >>> > > > proposed fix if I can identify the cause. > >>> > > > > >>> > > > I’ll have good availability over the next two weeks, so feel free > >>> to > >>> > tag > >>> > > me > >>> > > > on any follow-ups where I can help. > >>> > > > > >>> > > > Best Regards, > >>> > > > > >>> > > > Shilun Fan > >>> > > > > >>> > > > On Fri, Feb 13, 2026 at 6:33 AM Chris Nauroth < > [email protected] > >>> > > >>> > > wrote: > >>> > > > > >>> > > > > Hello everyone, > >>> > > > > > >>> > > > > I have bulk-moved remaining open 3.5.0 JIRA issues into a new > >>> 3.5.1 > >>> > > > > release. > >>> > > > > > >>> > > > > We have one remaining 3.5.0 blocker: HADOOP-19811. This has a > fix > >>> > > > available > >>> > > > > and a non-binding +1. > >>> > > > > > >>> > > > > https://github.com/apache/hadoop/pull/8243 > >>> > > > > > >>> > > > > Once a committer approves this, I'll proceed with branching and > >>> the > >>> > > rest > >>> > > > of > >>> > > > > the release process. > >>> > > > > > >>> > > > > We seem to have some test failures on trunk at the moment: > >>> > > > > > >>> > > > > > >>> https://github.com/apache/hadoop/pull/8237#issuecomment-3891386033 > >>> > > > > > >>> > > > > I haven't had a chance to investigate the cause yet, so I don't > >>> know > >>> > if > >>> > > > > these are going to be blockers. Any help there would be > >>> appreciated. > >>> > > > > > >>> > > > > Chris Nauroth > >>> > > > > > >>> > > > > >>> > > > >>> > > >>> > >> >
