upstart service unreliable after introducion of pipe to logger
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
tripleo |
Fix Released
|
Critical
|
Gregory Haynes |
Bug Description
upstart scripts for the openstack services now pipe the output to logger for example neutron-server the start script is:
The | introduces a fork which can in some circumtances leave a process behind after shutting down. For example after two runs of o-r-c I see
Oct 23 12:24:59 overcloud-
Oct 23 12:24:59 overcloud-
Oct 23 12:24:59 overcloud-
Oct 23 12:25:00 overcloud-
And then
Oct 23 12:27:07 overcloud-
Oct 23 12:27:07 overcloud-
Oct 23 12:27:08 overcloud-
Oct 23 12:27:09 overcloud-
But if I look at the system I see
root@overcloud-
neutron 8127 62.0 0.9 232516 74088 ? R 12:24 38:00 /opt/stack/
root 11370 0.0 0.0 4312 616 ? Ss 13:25 0:00 /bin/sh -e -c exec start-stop-daemon --start -c neutron --exec /opt/stack/
neutron 11371 14.5 1.5 286736 128432 ? S 13:25 0:04 /opt/stack/
root 11372 0.0 0.0 4212 340 ? S 13:25 0:00 logger -t neutron-server
root 11759 0.0 0.0 10464 864 pts/0 S+ 13:26 0:00 grep neutron-server
Now the process 8127 holds the neutron api server port 9696 so the second instance of neutron-server (11371) can bind to it and so goes into a infinite loop trying to start up.
The pipe was introduced in the review: https:/
Changed in tripleo: | |
status: | Fix Committed → Fix Released |
Fix proposed to branch: master /review. openstack. org/130830
Review: https:/