----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/21961/#review44180 -----------------------------------------------------------
Ship it! LGTM. src/master/master.cpp <https://reviews.apache.org/r/21961/#comment78499> Note that we are now removing the (non-checkpointing) frameworks after removing/rescinding the offers, rather than the other way around. I think this should be fine, since that slave is already disabled in the allocator, so no new offers will be made in the meantime. Since both removeFramework() and disconnect(Slave) remove offers, it shouldn't matter which one tries first. src/tests/fault_tolerance_tests.cpp <https://reviews.apache.org/r/21961/#comment78498> I think you can move this EXPECT_CALL down to just before detector.appoint(), since that is what triggers the authenticate/reregister that will do the disconnect/rescind. We don't want to confuse the reader into thinking that driver.start/register will rescind offers, since an unregistered slave will never get disconnected on authenticate. - Adam B On May 27, 2014, 7:14 p.m., Vinod Kone wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/21961/ > ----------------------------------------------------------- > > (Updated May 27, 2014, 7:14 p.m.) > > > Review request for mesos, Adam Berry, Benjamin Hindman, and Ben Mahler. > > > Bugs: MESOS-1418 > https://issues.apache.org/jira/browse/MESOS-1418 > > > Repository: mesos-git > > > Description > ------- > > See summary. > > > Diffs > ----- > > src/master/master.hpp f0490ec639d5c1f73ee5754bcca445daa9499caf > src/master/master.cpp 21472e94658be6e31f13eacc5412144d62b12096 > src/tests/fault_tolerance_tests.cpp > 3f2639347c7fde1b1b8e3c976e493b5fa2c3ade2 > > Diff: https://reviews.apache.org/r/21961/diff/ > > > Testing > ------- > > make check > > > Thanks, > > Vinod Kone > >
