> On Aug. 3, 2016, 7:58 p.m., Adam B wrote: > > src/master/master.cpp, lines 5164-5166 > > <https://reviews.apache.org/r/50723/diff/1/?file=1460860#file1460860line5164> > > > > Maybe increment `metrics->invalid_status_updates` in the else case here? > > And log here, as before?
This was pointed out by Neil earlier too. Continuing the discussion here. To me, it looked like an oversight in the prior code to treat status updates from agents for frameworks that have not yet re-registered upon a master failover as invalid and increment the metric? - Anand ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/50723/#review144662 ----------------------------------------------------------- On Aug. 3, 2016, 8:44 p.m., Anand Mazumdar wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/50723/ > ----------------------------------------------------------- > > (Updated Aug. 3, 2016, 8:44 p.m.) > > > Review request for mesos and Vinod Kone. > > > Bugs: MESOS-5930 > https://issues.apache.org/jira/browse/MESOS-5930 > > > Repository: mesos > > > Description > ------- > > The master's status handler function used to ignore the status updates > from the agents for frameworks not yet re-connected with the master > upon a failover. This change modifies that logic to still update > the local state and not bail out early. > > > Diffs > ----- > > src/master/master.cpp 060dc7f9730808c7fd9b8f9ecdbde0aac14d135c > src/tests/master_tests.cpp 6709818d599c068c289bcb714446018577082d8b > > Diff: https://reviews.apache.org/r/50723/diff/ > > > Testing > ------- > > make check (gtest_repeat=100) > > > Thanks, > > Anand Mazumdar > >
