Bug #2550
logrotate: SIGHUP upstart jobs too, not just sysvinit
History
#1 Updated by Anonymous over 11 years ago
The instance jobs make this a bit trickier. Either process "initctl list" output or copy the logic that walks the /var/lib/ceph/{mon,osd,...}/* directories and figures out cluster= and id= from the pathname. Must ignore "initctl reload" errors, to handle jobs exiting at an inopportune moment.
#2 Updated by Anonymous over 11 years ago
Filed upstream: https://bugs.launchpad.net/upstart/+bug/1012938
#3 Updated by Sage Weil over 11 years ago
- Assignee set to Sage Weil
- Priority changed from Normal to Urgent
- Target version set to v0.49
#4 Updated by Sage Weil over 11 years ago
- Status changed from New to Fix Under Review
Sigh. See branch upstart-vs-logrotate.
#5 Updated by Anonymous over 11 years ago
That killall thing is hideous, and I'm utterly unconvinced having even more upstart jobs for Ceph is helpful in any way.
Why is this not an initctl list|mawk loop straight in the logrotate config file?
#6 Updated by Sage Weil over 11 years ago
yeah, that'll work. only solves the logrotate case, but that's fine by me.
#7 Updated by Sage Weil over 11 years ago
repushed upstart-vs-logrotate branch
#8 Updated by Anonymous over 11 years ago
Please mention https://bugs.launchpad.net/upstart/+bug/1012938 in the "sucks" comment, so someone can some day nicely stumble on it once Upstart has improved.
I don't think osd actually uses "done" as a completion flag; ceph-disk-activate makes me think the word there is "active". Hrrmph.
I still thinking working from initctl list|mawk would have been better, it would also avoid spreading knowledge of the completion flag to this thing.
#9 Updated by Sage Weil over 11 years ago
- Status changed from Fix Under Review to Resolved