I'm afraid I haven't been paying enough attention to the jobs. Let's leave it as is for the moment and see. It would be nice to get the known flakey tests fixed or turned off though. Preference would be to fix.
Patrick On Fri, Mar 4, 2016 at 10:20 AM, Chris Nauroth <[email protected]> wrote: > I looked into ZooKeeper_branch35_jdk7 build #434, where all tests failed. > That one is very bizarre. There are numerous socket connection errors > coming from within RMI code, maybe JMX-related. The stack traces don't > even show any stack frames inside ZooKeeper code. Maybe we write this one > off as a transient environmental problem and keep an eye out? If it > happens again, maybe it's specific to a particular build host? > > Reviewing some other job runs, I couldn't find anything that would > specifically point to concurrent test execution as root cause. Failures > like ReconfigRecoveryTest (and the equivalent reconfig CPPUnit tests) and > FLETest are known issues documented in a few open JIRAs like > ZOOKEEPER-2080 and ZOOKEEPER-2164. We saw intermittent failures from > these even before we were running tests concurrently. > > If you do suspect concurrent test execution is a contributing factor, then > I think it would be fine to try turning it off for these jobs. The main > motivation for concurrent test execution was to improve the experience for > contributors by speeding up pre-commit. For these background polling > jobs, I think speed is a little less important. This can be reconfigured > entirely in the Jenkins jobs (no ZooKeeper build script changes required). > > --Chris Nauroth > > > > > On 3/4/16, 8:00 AM, "Patrick Hunt" <[email protected]> wrote: > >>Here's one where all the tests have failed, which is pretty unusual >>(threads is set to 8 on this job as well): >>https://builds.apache.org/view/S-Z/view/ZooKeeper/job/ZooKeeper_branch35_j >>dk7/434/ >> >>Patrick >> >>On Fri, Mar 4, 2016 at 7:36 AM, Patrick Hunt <[email protected]> wrote: >>> Hm, I'm seeing weird flakey tests, tests that typically haven't been >>> flakey. I also notice test.junit.threads is set to 8 on these jobs, >>> ala: https://issues.apache.org/jira/browse/ZOOKEEPER-2183 >>> >>> Here's one example >>> >>>https://builds.apache.org/view/S-Z/view/ZooKeeper/job/ZooKeeper_branch35_ >>>jdk8/1/ >>> >>> Chris would you be able to take a look? I suspect something in the >>> test setup with concurrent threads. >>> >>> Patrick >>> >>> On Thu, Mar 3, 2016 at 8:48 PM, Patrick Hunt <[email protected]> wrote: >>>> Also the flakey tests on jenkins are really painful. It's seems like >>>> it's gotten worse of late and it's more flakey than usual. Just me? >>>> Notice that we're missing a number of "standard" jobs for branch 3.5, >>>> I'll try to get the "standard" jobs setup asap. >>>> >>>> Patrick >>>> >>>> On Thu, Mar 3, 2016 at 4:28 PM, Patrick Hunt <[email protected]> wrote: >>>>> Looks like Flavio reopened this one, it's another license issue - >>>>> needs to get into 3.5.2 and trunk: >>>>> >>>>> https://issues.apache.org/jira/browse/ZOOKEEPER-2235 >>>>> >>>>> Patrick >>>>> >>>>> On Thu, Mar 3, 2016 at 3:56 PM, Patrick Hunt <[email protected]> wrote: >>>>>> FYI you can see my ZooKeeper Jira dashboard here: >>>>>>http://bit.ly/1UCURuM >>>>>> >>>>>> It's pretty useful to see overall project status as well as the in >>>>>> progress releases. >>>>>> >>>>>> Patrick >>>>>> >>>>>> On Thu, Mar 3, 2016 at 3:04 PM, Crowder Tim >>>>>><[email protected]> wrote: >>>>>>> Hi Chris- >>>>>>> >>>>>>> Any chance of getting ZOOKEEPER-1525 in there? >>>>>>> >>>>>>> Thanks! >>>>>>> .timrc >>>>>>> >>>>>>> -------------------------------------------- >>>>>>> On Thu, 3/3/16, Chris Nauroth <[email protected]> wrote: >>>>>>> >>>>>>> Subject: 3.5.2-alpha Release Planning >>>>>>> To: "[email protected]" <[email protected]> >>>>>>> Date: Thursday, March 3, 2016, 10:12 AM >>>>>>> >>>>>>> This link shows issues currently >>>>>>> fixed for 3.5.2: >>>>>>> >>>>>>> https://s.apache.org/n2C6 >>>>>>> >>>>>>> This link shows issues still open targeted to 3.5.2: >>>>>>> >>>>>>> https://s.apache.org/nORc >>>>>>> >>>>>>> I had been planning to kick off the release process for a >>>>>>> 3.5.2-alpha >>>>>>> candidate right around now, but after reviewing these lists, >>>>>>> I'm inclined >>>>>>> to defer it a few more weeks while we work down more >>>>>>> blockers. What do >>>>>>> others think? >>>>>>> >>>>>>> On a related note, thank you to Patrick for the recent JIRA >>>>>>> clean-up. >>>>>>> >>>>>>> --Chris Nauroth >> >
