On 10/12/15 11:27, "Michael Dürig" <mdue...@apache.org> wrote:
>> I’ve noticed that the failed builds can be assigned to a few >> machines. For instance, builds on ubuntu-6, ubuntu-5, H11 and H10 >> fails almost every time. On the other hand, all builds on the >> jenkins-ubuntu-1404-4gb succeedes. > >Good find! Are these machines slower in general (i.e. for all builds) or >only for the DocumentRDB builds? In the former case it might make sense >to just increase the time out. During the failing build tests are much slower (eg. AtomicCounterTest takes 82s in #604 rather than 4s in #603). I think there’s something specific about these machines that triggers a performance issue in the RDB or Derby. Speed of SegmentMK tests is comparable on all machines. >> For the further diagnostics it could be useful to get the >> unittest.log file from the failed build. Right now, the only way to >> do it is to wait until some build fails and grab the file from the >> workspace. The workspace is purged every time the jobs starts, so it >> requires a good timing. Maybe we can change the configuration, so the >> target/unittest.log file will be archived if the job is failed? > >This would be useful anyway but I think so far no one has figured out >how to do it. Unfortunately I can only grant committers access to that >Jenkins. Maybe Tommaso could help at this point as he was involved in >setting that instance up. Bertrand already pointed out the appropriate option. Passing **/target/unit-tests.log as the “Files to archive” should do the trick. Best regards, Tomek