Hi, I fixed some CI issues today.
https://issues.apache.org/jira/browse/ZEPPELIN-1583 https://issues.apache.org/jira/browse/ZEPPELIN-1857 https://issues.apache.org/jira/browse/ZEPPELIN-1986 https://issues.apache.org/jira/browse/ZEPPELIN-1749 Now, making CI green on PR wouldn't be too difficult. But we still have many flaky tests https://issues.apache.org/jira/browse/ZEPPELIN-1981. To keep CI stable, I think all committers should take care - Merge PR after pass the CI. - When CI fails with known flaky test, make sure each test profile pass at least once in multiple CI build attempts. Otherwise, new PR might break CI even though it looks perfectly okay in code review. Because there're cases that very difficult to catch in code review. For example, - New feature prints some log, and amount of log breaks CI. (CI log limit) - New feature increase start up time, and timeout in existing testing code all affected - New test consumes some memory to be run, which exceed memory of build machine. - .... Thanks, moon On Fri, Jan 20, 2017 at 12:20 AM Jongyoul Lee <jongy...@gmail.com> wrote: > I can change some fields and types but cannot convert all tickets to > sub-task. I'll change them manually. > > On Thu, Jan 19, 2017 at 4:20 PM, Jeff Zhang <zjf...@gmail.com> wrote: > > > Hi Jongyoul, > > > > Could you move all these tickets under this umbrella ticket ? > > > > You can do bulk change by clicking the gear tool button on top right of > > this page. ( I don't' have privilege to do that) > > > > > > https://issues.apache.org/jira/issues/?jql=project%20% > > 3D%20ZEPPELIN%20AND%20text%20~%20flaky%20and%20status%20!% > > 3D%20RESOLVED%20ORDER%20%20BY%20status%20ASC%20 > > > > > > > > > > Jongyoul Lee <jongy...@gmail.com>于2017年1月19日周四 下午12:11写道： > > > > > Thanks Jeff, > > > > > > I made a new PR for handling this issue. > > > > > > https://issues.apache.org/jira/browse/ZEPPELIN-1981 > > > > > > On Thu, Jan 19, 2017 at 9:47 AM, Jianfeng (Jeff) Zhang < > > > jzh...@hortonworks.com> wrote: > > > > > > > +1 > > > > > > > > Best Regard, > > > > Jeff Zhang > > > > > > > > > > > > > > > > > > > > > > > > On 1/19/17, 1:03 AM, "Jongyoul Lee" <jongy...@gmail.com> wrote: > > > > > > > > >I also agree that ppl don't care of the result of CI anymore even > it's > > > > >real > > > > >failure. One possible solution is making umbrella ticket, grabbing > > flaky > > > > >tests, disabling at first and enabling when it solves. but it > assumes > > we > > > > >need to do our best to fix the flaky tests. Otherwise, we will lose > > some > > > > >tests... > > > > > > > > > >How do you guys think of it? > > > > > > > > > >On Thu, Jan 19, 2017 at 1:56 AM, Felix Cheung < > > > felixcheun...@hotmail.com> > > > > >wrote: > > > > > > > > > >> I'd agree. Is there a course of actions you can propose? Disable > all > > > > >>these > > > > >> tests is a not a long term solution, right? > > > > >> > > > > >> > > > > >> _____________________________ > > > > >> From: Jeff Zhang <zjf...@gmail.com<mailto:zjf...@gmail.com>> > > > > >> Sent: Tuesday, January 17, 2017 6:01 PM > > > > >> Subject: Re: Unstable travis CI recently > > > > >> To: <dev@zeppelin.apache.org<mailto:dev@zeppelin.apache.org>> > > > > >> > > > > >> > > > > >> > > > > >> Should we disable these flaky test now ? CI seems become more > > unstable > > > > >> recently. It is almost useless for me, I never see a success CI > > > > >>recently. > > > > >> Here's one screenshot of recent closed PRs. Most of them has CI > > > failure. > > > > >> IMO, this is pretty bad, especially for new contributors. > > > > >> > > > > >> > > > > >> > > > > >> [pasted1] > > > > >> > > > > >> > > > > >> > > > > >> > > > > >> > > > > >> Jun Kim <i2r....@gmail.com<mailto:i2r....@gmail.com > >>?2016?12?13??? > > > > >> ??11:27??? > > > > >> @Hoon Thanks for your information :-) I should use that next time! > > > > >> > > > > >> 2016? 12? 13? (?) ?? 8:23, Park Hoon <1am...@gmail.com<mailto: > 1amb4 > > > > >> a...@gmail.com>>?? ??: > > > > >> > > > > >> > I totally agree with your opinions. I will work on > ZEPPELIN-1739, > > > > >> > ZEPPELIN-1749 first i reported before. > > > > >> > > > > > >> > @Jun Kim. So true. We have to wait long time. FYI, we can use > our > > > own > > > > >> > travis CI containers to test (I recently learned also!) by > > > configuring > > > > >> > your-github-id/zeppelin-repo in travis CI > > > > >> > > > > > >> > Thanks! > > > > >> > > > > > >> > On Tue, Dec 13, 2016 at 8:19 PM, Jun Kim > > > > >><i2r....@gmail.com<mailto:i2r. > > > > >> j...@gmail.com>> wrote: > > > > >> > > > > > >> > > I definitely agree with you! > > > > >> > > > > > > >> > > I reopened my PR twice recently to pass CI and it wasn't > because > > > of > > > > >>me. > > > > >> > > > > > > >> > > CI takes about ~40min for a test, so I had to wait 1h and > 20min > > to > > > > >> write > > > > >> > a > > > > >> > > comment after passing CI T_T > > > > >> > > > > > > >> > > And the worst of it is that I don't believe CI's result more > and > > > > >>more. > > > > >> > > > > > > >> > > 2016? 12? 13? (?) ?? 8:10, Jeff Zhang <zjf...@gmail.com > <mailto: > > > > zjffd > > > > >> u...@gmail.com>>?? ??: > > > > >> > > > > > > >> > > > Hi Folks, > > > > >> > > > > > > > >> > > > As you may notice that our travis CI is not stable recently. > > > > >>There's > > > > >> > many > > > > >> > > > flaky test, and it waste every developer's time to figure > out > > > > >>whether > > > > >> > the > > > > >> > > > failure is due to your PR or flaky test. So I think it is > time > > > > >>for us > > > > >> > to > > > > >> > > > make the CI stable. Here's tickets for all the flaky test. > > > > >> > > > > > > > >> > > > https://issues.apache.org/jira/issues/?jql=project%20% > > > > >> > > 3D%20ZEPPELIN%20AND%20text%20~%20flaky%20and%20status%20!% > > > > >> > > 3D%20RESOLVED%20ORDER%20%20BY%20status%20ASC%20 > > > > >> > > > > > > > >> > > > Fixing the flaky test may take some time and may not easy > for > > > some > > > > >> test > > > > >> > > but > > > > >> > > > I think it is worth to do that. And it is better for these > > > people > > > > >> who > > > > >> > > > familiar with that particular test case to fix it. What do > you > > > > >>guys > > > > >> > > think ? > > > > >> > > > Thanks > > > > >> > > > > > > > >> > > -- > > > > >> > > Taejun Kim > > > > >> > > > > > > >> > > Data Mining Lab. > > > > >> > > School of Electrical and Computer Engineering > > > > >> > > University of Seoul > > > > >> > > > > > > >> > > > > > >> -- > > > > >> Taejun Kim > > > > >> > > > > >> Data Mining Lab. > > > > >> School of Electrical and Computer Engineering > > > > >> University of Seoul > > > > >> > > > > >> > > > > >> > > > > > > > > > > > > > > >-- > > > > >이종열, Jongyoul Lee, 李宗烈 > > > > >http://madeng.net > > > > > > > > > > > > > > > > > -- > > > 이종열, Jongyoul Lee, 李宗烈 > > > http://madeng.net > > > > > > > > > -- > 이종열, Jongyoul Lee, 李宗烈 > http://madeng.net >
