> On June 14, 2013, 2:24 a.m., Ben Mahler wrote: > > src/linux/cgroups.cpp, lines 1515-1518 > > <https://reviews.apache.org/r/11131/diff/5/?file=304997#file304997line1515> > > > > On second thought, let's just keep it simple for now as I imagine it's > > alright to freeze an empty cgroup. > > > > So stop -> kill -> empty -> freeze -> kill -> thaw -> empty. Then we > > don't need to be concerned with only continuing the chain when the kill > > doesn't work. > > > > Disregard SIGCONT, I was mislead by some comments in killtree.sh but > > it's not necessary when using SIGKILL. > > > > I apologize for the conflicting feedback!! To be clear, we haven't made > > any progress either in terms of figuring out what the problem is so we're > > agreed on this fix! :) > > Vinod Kone wrote: > can't we just do stop->kill->freeze->kill->thaw->empty?
Ben? Thoughts on this? - Brenden ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/11131/#review21893 ----------------------------------------------------------- On June 17, 2013, 7:22 p.m., Brenden Matthews wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/11131/ > ----------------------------------------------------------- > > (Updated June 17, 2013, 7:22 p.m.) > > > Review request for mesos. > > > Description > ------- > > Changed cgroups killTasks() sequence. > > The sequence is as follows: > stop -> kill -> empty -> freeze -> kill -> thaw -> empty > > We also now ignore ERSCH errors from kill() in cgroups::kill(). > > Review: https://reviews.apache.org/r/11131 > > > Diffs > ----- > > src/linux/cgroups.cpp 8d94fe63610c4c7a48f92d260fcd526b7a83942e > > Diff: https://reviews.apache.org/r/11131/diff/ > > > Testing > ------- > > Used in production at airbnb. > > make -j10 check && cd hadoop && make hadoop-2.0.0-mr1-cdh4.2.1 && make > hadoop-0.20.205.0 && make hadoop-0.20.2-cdh3u3 > > > Thanks, > > Brenden Matthews > >
