Jason: I wouldn’t object to reverting immediately. It’s a courtesy to give someone a chance to fix it themselves rather than unilaterally revert is all.
> On Sep 18, 2020, at 3:05 PM, Tomás Fernández Löbbe <[email protected]> > wrote: > > Sometimes Jenkins may take hours to take your commit, may fail in the middle > of your night, may not fail consistently, etc. That's why I don't think > giving specific timeframes makes sense, but yes, as soon as you notice it's > failing, it's either fix immediately or revert IMO. > > On Fri, Sep 18, 2020 at 12:03 PM Jason Gerlowski <[email protected]> > wrote: > > If it’s inadvertently added, we either fix it within an hour or so or > > revert the offending commit > > > I don't want to set specific time frames, > > To play Devil's Advocate here: why wait even an hour to revert a 100% > test failure? Reverts are usually trivial to do, unblock others > immediately, and don't interfere with the fix process at all. > Remembering the times I've broken the build myself, reverts even seem > preferable from that position - reverting up front takes all the > time-pressure off of getting out a fix. Why work under the gun when > you don't have to? > > On Fri, Sep 18, 2020 at 1:14 PM Tomás Fernández Löbbe > <[email protected]> wrote: > > > > I believe these failures are associated to > > https://issues.apache.org/jira/browse/SOLR-14151 > > > > • FAILED: org.apache.solr.pkg.TestPackages.classMethod > > • FAILED: > > org.apache.solr.schema.PreAnalyzedFieldManagedSchemaCloudTest.testAdd2Fields > > • FAILED: > > org.apache.solr.schema.ManagedSchemaRoundRobinCloudTest.testAddFieldsRoundRobin > > > > > IMO if a temporary instability is to be introduced deliberately, it > > > should be published on the list. If it’s inadvertently added, we either > > > fix it within an hour or so or revert the offending commit > > I don't want to set specific time frames, but sometimes it's obviously too > > much time. > > > > On Fri, Sep 18, 2020 at 8:48 AM Atri Sharma <[email protected]> wrote: > >> > >> When I said temporary, I meant 3-4 hours. Definitely not more than that. > >> > >> IMO we should just roll back offending commits if they are easily > >> identifiable. I agree with you — we all have been guilty of breaking > >> builds (mea culpa as well). The bad part here is the longevity of the > >> failures. > >> > >> > >> On Fri, 18 Sep 2020 at 21:05, Erick Erickson <[email protected]> > >> wrote: > >>> > >>> bq. IMO if a temporary instability is to be introduced deliberately, it > >>> should be published on the list > >>> > >>> > >>> > >>> Actually, I disagree. Having anything in the tests that fail 100% of the > >>> time is just unacceptable since it becomes a barrier for everyone else. > >>> AFAIK, if the problem can be identified to a particular push, I have no > >>> problems with that push being unilaterally rolled back. > >>> > >>> > >>> > >>> The exception for me is when the problem is addressed immediately, I’ve > >>> certainly been the source of that kind of problem, as have others. > >>> > >>> > >>> > >>> What I take great exception to is the fact that some of these tests have > >>> been failing 100% of the time for the last seven days! If it’s the case > >>> that the full test suite was never run before the push that’s another > >>> discussion. Yeah, it takes a long time but… > >>> > >>> > >>> > >>> Erick > >>> > >>> > >>> > >>> > On Sep 18, 2020, at 11:28 AM, Atri Sharma <[email protected]> wrote: > >>> > >>> > > >>> > >>> > IMO if a temporary instability is to be introduced deliberately, it > >>> > should be published on the list. If it’s inadvertently added, we either > >>> > fix it within an hour or so or revert the offending commit. > >>> > >>> > > >>> > >>> > On Fri, 18 Sep 2020 at 20:26, Erick Erickson <[email protected]> > >>> > wrote: > >>> > >>> > http://fucit.org/solr-jenkins-reports/failure-report.html > >>> > >>> > > >>> > >>> > > >>> > >>> > > >>> > >>> > HdfsAutoAddReplicasTest failing 100% of the time. > >>> > >>> > > >>> > >>> > TestPackages.classMethod failing 100% of the time > >>> > >>> > > >>> > >>> > 3-4 AutoAddReplicas tests failing 98% of the time. > >>> > >>> > > >>> > >>> > > >>> > >>> > > >>> > >>> > Is anyone looking at these? I realize the code base is changing a lot, > >>> > and some temporary instability is to be expected. What I’d like is for > >>> > some indication that people are actively addressing these. > >>> > >>> > > >>> > >>> > > >>> > >>> > > >>> > >>> > Erick > >>> > >>> > > >>> > >>> > --------------------------------------------------------------------- > >>> > >>> > > >>> > >>> > To unsubscribe, e-mail: [email protected] > >>> > >>> > > >>> > >>> > For additional commands, e-mail: [email protected] > >>> > >>> > > >>> > >>> > > >>> > >>> > > >>> > >>> > -- > >>> > >>> > Regards, > >>> > >>> > > >>> > >>> > Atri > >>> > >>> > Apache Concerted > >>> > >>> > >>> > >>> > >>> > >>> --------------------------------------------------------------------- > >>> > >>> To unsubscribe, e-mail: [email protected] > >>> > >>> For additional commands, e-mail: [email protected] > >>> > >>> > >>> > >> -- > >> Regards, > >> > >> Atri > >> Apache Concerted > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
