> On June 15, 2019, 5:34 p.m., Benjamin Mahler wrote: > > Thanks! One thing I'm wondering is whether it makes sense to just populate > > `FrameworkInfo.user` and `FrameworkInfo.hostname` in > > `SchedulerProcess::updateFramework` to the previous values (if not set), > > rather than calling the system calls again.
I'm not sure either. This approach has two upsides: - a bit simpler - calling updateFramework() is consistent with re-creating the SchedulerDriver and two downsides: - user/hostname can change during the SchedulerDriver's lifetime (weird) - these syscalls might block for an enormous length of time (at least the ones behind `net::hostname()`, which calls `getaddrinfo()`, which might perform DNS lookups) - Andrei ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/70854/#review215916 ----------------------------------------------------------- On June 14, 2019, 12:32 p.m., Andrei Sekretenko wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/70854/ > ----------------------------------------------------------- > > (Updated June 14, 2019, 12:32 p.m.) > > > Review request for mesos and Benjamin Mahler. > > > Repository: mesos > > > Description > ------- > > This patch makes it mandatory for the caller to fill the 'id' field of > the FrameworkInfo passed into MesosSchedulerDriver::updateFramework(). > > > Diffs > ----- > > include/mesos/scheduler.hpp c5e61d4fd7aae5067d8c7d8dec878dc82c86e4ec > src/sched/sched.cpp 281236bb53e1c6ed77b69bf954e27705595ffb2a > > > Diff: https://reviews.apache.org/r/70854/diff/2/ > > > Testing > ------- > > > Thanks, > > Andrei Sekretenko > >
