----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/71297/#review217239 -----------------------------------------------------------
src/tests/operation_reconciliation_tests.cpp Lines 1842 (patched) <https://reviews.apache.org/r/71297/#comment304522> We are restarting the master once, but the scheduler gets connected/disconnected event pair twice during the master restart... is this guaranteed by TestMesos + StandaloneMasterDetector, a stable coincidence, or just the most likely outcome of some other race? IMO, this is at least worth a comment - but what about preventing it? (Will something like `detector->appoint(None())` + `AWAIT_READY(disconnected)` before killing the master help?) - Andrei Sekretenko On Aug. 16, 2019, 1:57 p.m., Benno Evers wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/71297/ > ----------------------------------------------------------- > > (Updated Aug. 16, 2019, 1:57 p.m.) > > > Review request for mesos, Andrei Sekretenko, Greg Mann, Joseph Wu, and Till > Toenshoff. > > > Bugs: MESOS-9928 > https://issues.apache.org/jira/browse/MESOS-9928 > > > Repository: mesos > > > Description > ------- > > The FrameworkReconciliationRaceWithUpdateSlave test from the > operation reconciliation tests was flaky since we did not wait > for the scheduler to reconnect before advancing the clock to > trigger reregistration. > > > Diffs > ----- > > src/tests/operation_reconciliation_tests.cpp > 9d084c027ec2f910515cafebf715f7428c43f1a9 > > > Diff: https://reviews.apache.org/r/71297/diff/1/ > > > Testing > ------- > > `./src/mesos-tests > --gtest_filter="*FrameworkReconciliationRaceWithUpdateSlaveMessage*" > --gtest_repeat=200` while simultaneously running `stress-ng` in the > background. > > > Thanks, > > Benno Evers > >