----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/25947/#review54293 -----------------------------------------------------------
3rdparty/libprocess/src/reap.cpp <https://reviews.apache.org/r/25947/#comment94305> nit: it's easy for someone to change the constants in the rest of this calculation but not reconfigure this linear relationship. It would be better to pull out 50, Milliseconds(100), 500, and Seconds(1) as constants then calculate this linear relationship using those. - Dominic Hamon On Sept. 23, 2014, 11:05 a.m., Ian Downes wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/25947/ > ----------------------------------------------------------- > > (Updated Sept. 23, 2014, 11:05 a.m.) > > > Review request for mesos, Bernd Mathiske, Ben Mahler, Craig Hansen-Sturm, and > Jie Yu. > > > Bugs: MESOS-1199 > https://issues.apache.org/jira/browse/MESOS-1199 > > > Repository: mesos-git > > > Description > ------- > > If pid count <= 50 then use 100 ms (<= 0.5% core usage), if count >= 500 use > 1000 ms (<= 1% core usage at 500 pids), else interpolate. > > > Diffs > ----- > > 3rdparty/libprocess/src/reap.cpp b350ee15bf232493be01506037524b7590c0d2f5 > > Diff: https://reviews.apache.org/r/25947/diff/ > > > Testing > ------- > > Wrote test program which looped, calling kill(0, pid) and waitpid(pid, > &status, WNOHANG) on N children (still running) each loop. These are the > system calls used by the reaper. Computed load as wall time / (user + system > time). Tested on Linux and OSX. > > `sudo ./bin/mesos-tests.sh` time reduced 20% from 6:40 to 5:20 (not running > Docker tests). > > > Thanks, > > Ian Downes > >
