> On Oct. 21, 2015, 12:41 a.m., Joris Van Remoortere wrote: > > Hey Mandeep, Thanks for taking on this important work! > > Some high level questions: > > - Do we need to do reference counting manually or can we use a construct > > like `Shared` in Libprocess that wraps the `std::shared_ptr`. > > - Do we want to surface the singleton pattern in this way? What are the > > impacts of this on the user (i.e. your logic calling the `acquire` / > > `release` code currently? Does the user care about holding a reference to > > the singleton? > > > > Regardless, I added some style comments as they are valuable for you as a > > contributor.
Thanks Joris for the comments. Let me get that done and upload the new diff. - Mandeep ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/39060/#review103331 ----------------------------------------------------------- On Oct. 8, 2015, 5:05 p.m., Mandeep Chadha wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/39060/ > ----------------------------------------------------------- > > (Updated Oct. 8, 2015, 5:05 p.m.) > > > Review request for mesos and Joris Van Remoortere. > > > Bugs: MESOS-3595 > https://issues.apache.org/jira/browse/MESOS-3595 > > > Repository: mesos > > > Description > ------- > > If the number of framework created exceeds the lib process > threads then during master failover the zookeeper updates can > cause deadlock. > > > Diffs > ----- > > src/sched/sched.cpp 571e00d303009a940f17c8ed4582749a718e846d > > Diff: https://reviews.apache.org/r/39060/diff/ > > > Testing > ------- > > make check successful > Created 100 framework instances on a 24 CPU machine. Master failover detected > by the framework process and continue to work as expected. > > > Thanks, > > Mandeep Chadha > >
