Last modified: 2014-10-31 00:24:19 UTC
THere has been some discussion on irc of the runtime resource consumption of a "normal" MediaWiki-Vagrant install. It seems like time to spin up a brand new VM and do a process by process inspection of the resources that are being used with an eye to removing things that unnecessary.
Should be a fun evening of detective work.
Start a clean vm: * vagrant destroy -f * vagrant reset-roles * vagrant up * vagrant ssh Check resources in use: $ ps ax|wc -l 95 $ free -m total used free shared buffers cached Mem: 993 839 154 8 36 510 -/+ buffers/cache: 292 701 Swap: 0 0 0 Top memory use: 8.2% hhvm --mode=server 6.1% php /srv/jobrunner/redisJobRunnerService 5.4% mysql 2.1% apache 0.7% redis-server The ram consumed by the 2 idle hhvm containers is an obvious target to find improvements for. (The php that is running the jobrunner is just hhvm in disguise.) I have seen `puppet agent` running occasionally too, but not consistently.
Change 160908 had a related patch set uploaded by BryanDavis: Disable puppet service https://gerrit.wikimedia.org/r/160908
Change 160908 merged by jenkins-bot: Disable puppet service https://gerrit.wikimedia.org/r/160908
(In reply to Bryan Davis from comment #2) > > I have seen `puppet agent` running occasionally too, but not consistently. This should be fixed by https://gerrit.wikimedia.org/r/160908
Change 162631 had a related patch set uploaded by BryanDavis: Do not start puppet agent in Labs https://gerrit.wikimedia.org/r/162631
Change 162631 merged by jenkins-bot: Do not start puppet agent in Labs https://gerrit.wikimedia.org/r/162631
There is probably more that can be done on this front, but short of changing the jobrunner to use php5 instead of hhvm I don't see any easy wins.