> On Aug. 13, 2019, 1:10 p.m., Benno Evers wrote: > > src/master/master.cpp > > Lines 7959 (patched) > > <https://reviews.apache.org/r/71275/diff/1/?file=2160640#file2160640line7959> > > > > Can we already decide at this point if the agent is drained? E.g. > > `checkAndTransitionDrainingAgent()` seems to look at pending operations, > > which only get sent in the first `UpdateSlaveMessage` after the > > reregistration is completed.
Good catch Benno - this is also an issue with the existing reregistration code in `___reregisterSlave()`. Unfortunately it's even more annoying than waiting for the first `UpdateSlaveMessage`, since it's possible that resource providers may subscribe with the agent later on. Not yet sure what the best way to handle this is. - Greg ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/71275/#review217177 ----------------------------------------------------------- On Aug. 13, 2019, 2:35 a.m., Joseph Wu wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/71275/ > ----------------------------------------------------------- > > (Updated Aug. 13, 2019, 2:35 a.m.) > > > Review request for mesos, Benjamin Bannier, Benno Evers, and Greg Mann. > > > Bugs: MESOS-9934 > https://issues.apache.org/jira/browse/MESOS-9934 > > > Repository: mesos > > > Description > ------- > > This logic was missing from the initial implementation of agent > draining. When an agent became unreachable, and then reregistered > with the master, the master would not properly deactivate or drain > the agent. > > > Diffs > ----- > > src/master/master.cpp 31c7c97abecb92591369b417e2ef38a99cc09bac > src/tests/api_tests.cpp c2099674e742eaa08134c5e0a7cdab1734808119 > > > Diff: https://reviews.apache.org/r/71275/diff/1/ > > > Testing > ------- > > make check > > > Thanks, > > Joseph Wu > >
