> On Feb. 12, 2018, 12:35 p.m., Joseph Wu wrote: > > src/tests/default_executor_tests.cpp > > Lines 3450-3461 (patched) > > <https://reviews.apache.org/r/65552/diff/2/?file=1954220#file1954220line3450> > > > > Is it possible for the following race to occur? > > > > * Executor launches task group 1 (expected to fail/kill) > > * Executor performs the launch/kill. > > * Executor commits suicide because it is no longer running any tasks. > > * The agent sends the second task group to the now-dead executor.
Yeah, that sounds possible, I changed the test so that it now does the following: 1. Executor launches `taskGroup1` with a task that sleeps for a very long time and isn't expected to stop until killed. 2. Executor launches `taskGroup2` with a sleep task and one that should fail to launch. 3. Executor should kill the sleep task in `taskGroup2`. 4. Executor report all tasks in `taskGroup2` as killed/failed. 4. Scheduler will ask to kill the sole task in `taskGroup1`. 5. Executor should kill the task in `taskGroup1` and terminate. - Gaston ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/65552/#review197312 ----------------------------------------------------------- On Feb. 12, 2018, 3:24 p.m., Gaston Kleiman wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/65552/ > ----------------------------------------------------------- > > (Updated Feb. 12, 2018, 3:24 p.m.) > > > Review request for mesos, Anand Mazumdar, Greg Mann, Qian Zhang, and Vinod > Kone. > > > Bugs: MESOS-8468 > https://issues.apache.org/jira/browse/MESOS-8468 > > > Repository: mesos > > > Description > ------- > > Added a regression test for MESOS-8468. > > > Diffs > ----- > > src/tests/default_executor_tests.cpp > cc97e0d1fea7f4d0bc544d850593d8d91921b552 > > > Diff: https://reviews.apache.org/r/65552/diff/3/ > > > Testing > ------- > > `GLOG_v=1 sudo bin/mesos-tests.sh --gtest_filter='*ROOT_LaunchGroupFailure*' > --verbose --gtest_repeat=650 --gtest_break_on_failure` on GNU/Linux > > > Thanks, > > Gaston Kleiman > >
