----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/5402/ -----------------------------------------------------------
(Updated July 5, 2012, 10:07 p.m.) Review request for mesos, Benjamin Hindman and Vinod Kone. Changes ------- Refactor the code. Use .then to link several operations in TasksKill. Separate cgroup destroy code into a new patch (will submit later). Include a new test. Summary (updated) ----------------- Add an API to atomically kill all tasks in a cgroup. Description ------- This patch leverages the freezer subsystem in cgroups to kill all the processes in a cgroup atomically. The main idea is to freeze all the processes in a cgroup first, then send kill signal to all the proceses. This avoids the need of walking the proc process tree to kill all processes associated with an executor. In fact, the original killtree solution assumes that the user processes haven't blocked the SIGSTOP signal, which may not be true in some cases. Diffs (updated) ----- src/linux/cgroups.hpp PRE-CREATION src/linux/cgroups.cpp PRE-CREATION src/tests/cgroups_tests.cpp PRE-CREATION Diff: https://reviews.apache.org/r/5402/diff/ Testing ------- On Linux machine, make check. Thanks, Jie Yu
