----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/26200/#review55141 -----------------------------------------------------------
Ship it! src/master/master.cpp <https://reviews.apache.org/r/26200/#comment95524> How about: // Master expects ExecutorInfo.framework_id to be // set even though it is an optional field. Currently the // scheduler driver ensures that the field is set. // But for schedulers not using the driver, we need // to do the validation here. // TODO(bmahler): Set this field in the master instead // of depending on the scheduler driver do it. - Vinod Kone On Sept. 30, 2014, 11:30 p.m., Ben Mahler wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/26200/ > ----------------------------------------------------------- > > (Updated Sept. 30, 2014, 11:30 p.m.) > > > Review request for mesos and Vinod Kone. > > > Repository: mesos-git > > > Description > ------- > > We set this field on behalf of frameworks within the scheduler driver. > > When we move away from the scheduler driver, we'll want to consider setting > this field on behalf of the framework within the Master. However, for now, > since the master will crash if it's unset [1], we should prevent this from > occurring. > > [1] https://github.com/apache/mesos/blob/0.20.1/src/master/master.cpp#L4134 > > > Diffs > ----- > > src/master/master.cpp a60308f912a1ed81ecd51c677461a8f591d9eb8e > > Diff: https://reviews.apache.org/r/26200/diff/ > > > Testing > ------- > > make check, adding a new test for this requires intercepting and modifying a > LaunchTaskMessage, since the driver always sets this field. > > In the longer term, we should add unit tests for message validators. > > > Thanks, > > Ben Mahler > >