Job queue
(→Emergency kill) |
|||
| (10 intermediate revisions by 3 users not shown) | |||
| Line 1: | Line 1: | ||
| − | + | __TOC__ | |
| − | + | ==Overview== | |
| + | Job queue runners run on the job-runners node group. Install a new job runner using the application::jobrunner class in puppet. | ||
| − | + | The daemon is controlled via <tt>/etc/init.d/mw-job-runner</tt> and will start on boot by default. | |
| − | + | ||
| − | + | ||
See also http://www.mediawiki.org/wiki/Manual:Job_queue | See also http://www.mediawiki.org/wiki/Manual:Job_queue | ||
| + | |||
| + | ==Emergency kill== | ||
| + | If there's an urgent call to kill job queues everywhere for example because they're causing clusterwide swapdeath:<br /> | ||
| + | - on the job-runner group (see /home/config/others/usr/local/dsh/node_groups) . . .<br /> | ||
| + | - as root . . .<br /> | ||
| + | - dsh -g job-runners pkill -9 -f obs (this matches jobs-loop and RunJobs and few others, we don't trap sinals in our php scripts so there's no need to kill gracefully)<br /> | ||
| + | - if you end up having to power cycle hosts that are nonresponsive, remember to rerun the dsh afterward to make sure job-runners is stopped there as well until someone can investigate the memory issue | ||
Latest revision as of 15:55, 19 January 2012
Contents |
[edit] Overview
Job queue runners run on the job-runners node group. Install a new job runner using the application::jobrunner class in puppet.
The daemon is controlled via /etc/init.d/mw-job-runner and will start on boot by default.
See also http://www.mediawiki.org/wiki/Manual:Job_queue
[edit] Emergency kill
If there's an urgent call to kill job queues everywhere for example because they're causing clusterwide swapdeath:
- on the job-runner group (see /home/config/others/usr/local/dsh/node_groups) . . .
- as root . . .
- dsh -g job-runners pkill -9 -f obs (this matches jobs-loop and RunJobs and few others, we don't trap sinals in our php scripts so there's no need to kill gracefully)
- if you end up having to power cycle hosts that are nonresponsive, remember to rerun the dsh afterward to make sure job-runners is stopped there as well until someone can investigate the memory issue
