> On March 21, 2014, 2:46 p.m., Ian Downes wrote: > > src/launcher/executor.cpp, line 308 > > <https://reviews.apache.org/r/19239/diff/2/?file=520053#file520053line308> > > > > can you refactor to avoid this duplication? perhaps a killAll() method > > that takes a signal and is called by shutdown() with SIGTERM and then timer > > with SIGKILL.
Cool - good idea. Should all be done in the new patch :) - Niklas ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/19239/#review38213 ----------------------------------------------------------- On March 25, 2014, 11:56 a.m., Niklas Nielsen wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/19239/ > ----------------------------------------------------------- > > (Updated March 25, 2014, 11:56 a.m.) > > > Review request for mesos, Ben Mahler and Ian Downes. > > > Bugs: MESOS-1031 > https://issues.apache.org/jira/browse/MESOS-1031 > > > Repository: mesos-git > > > Description > ------- > > This patch adds a two step signal escalation: 1) SIGTERM is sent to > the process tree 2) If the process tree root hasn't been reaped within > escalationTimeout (3 seconds), SIGKILL is sent to the process tree. > Currently, this may leave behind orphan processes if parent processes > terminates and leave non-responsive processes behind. This will be > captured when PID namespaces are in place, as the orphan processes > will end up hanging off the process tree root instead of init. > > > Diffs > ----- > > src/launcher/executor.cpp e30d77a > src/slave/constants.hpp d237383 > src/slave/constants.cpp 3fb0cde > > Diff: https://reviews.apache.org/r/19239/diff/ > > > Testing > ------- > > make check > > > Thanks, > > Niklas Nielsen > >
