----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/46923/#review131635 -----------------------------------------------------------
I notice that `Master::_exited` also checks that the `failover_timeout` is valid. Do you think it would be better to simplify that logic to assume that the `failover_timeout` is always valid? i.e., add a `CHECK` that it is valid rather than printing a warning message. src/master/master.hpp (line 938) <https://reviews.apache.org/r/46923/#comment195657> You can remove this change. src/master/master.cpp (line 2256) <https://reviews.apache.org/r/46923/#comment195660> I think using `stringify` here would be a bit more idiomatic in Mesos. src/master/master.cpp (line 2479) <https://reviews.apache.org/r/46923/#comment195658> This should be consistent with the first instance of the error message above. src/tests/master_tests.cpp (line 4481) <https://reviews.apache.org/r/46923/#comment195659> This comment still seems unclear to me -- we should explain why `99999999999999999` is an invalid framework timeout. (The reason is that we internally represent the timeout using nanoseconds since the epoch as an int64; `99999999999999999` in seconds converted to nanoseconds is too large for an int64). Also, the conversion fails internally to Mesos (stout), not because of protobuf. - Neil Conway On May 4, 2016, 12:43 a.m., Jose Guilherme Vanz wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/46923/ > ----------------------------------------------------------- > > (Updated May 4, 2016, 12:43 a.m.) > > > Review request for mesos, Guangya Liu, Kevin Sweeney, Neil Conway, and Vinod > Kone. > > > Bugs: MESOS-1575 > https://issues.apache.org/jira/browse/MESOS-1575 > > > Repository: mesos > > > Description > ------- > > Frameworks were allowed to subscribe themselves with invalid failover > timeout. For this reason, a validation has been made in the master to > deny framework subscription if it set a invalid value for the failover > timeout. MESOS-1575 > > > Diffs > ----- > > src/master/master.hpp 3e55114ee7866e06513071e86e15608099dae052 > src/master/master.cpp 6d3e0f7c634690a35eec1ce827b705e04c3af87e > src/tests/master_tests.cpp 8e00753fcbcd0cae1d08aad387b08aa17c7f63ad > > Diff: https://reviews.apache.org/r/46923/diff/ > > > Testing > ------- > > > Thanks, > > Jose Guilherme Vanz > >
