logrotate script for apache leads to restarting keystone service
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mirantis OpenStack |
Fix Released
|
Critical
|
Sergii Golovatiuk | ||
7.0.x |
Fix Released
|
Critical
|
Sergii Golovatiuk | ||
8.0.x |
Fix Released
|
Medium
|
MOS Keystone |
Bug Description
during boot_server_
http://
Cinder can't download image from glance. from cinder-volume.log:
http://
determine glance server from haproxy:
http://
Glance get 500 from keystone when tries to check token. from glance log:
http://
Keystone restarting at the time. from keystone log:
http://
It happens every day at ~07:35. from /var/log/
http://
It seems logrotating is root cause of the behavior. /etc/logrotate.
http://
If we really need to restart keystone service, then we need to do that not at the same time on all controller nodes.
Cluster configuration:
Baremetal,
Controllers:3 Computes:180 Copmutes+Ceph: 20
api: '1.0'
astute_sha: e24ca066bf6160b
auth_required: true
build_id: 2015-08-17_03-04-59
build_number: '182'
feature_groups:
- mirantis
fuel-agent_sha: 57145b1d8804389
fuel-library_sha: 9de2625d26c3b88
fuel-nailgun-
fuel-ostf_sha: 17786b86b78e5b6
fuelmain_sha: d8c726645be087b
nailgun_sha: 4710801a2f4a6d6
openstack_version: 2015.1.0-7.0
production: docker
python-
release: '7.0'
Diagnostic Snapshot: http://
Changed in mos: | |
assignee: | Fuel Library Team (fuel-library) → Sergii Golovatiuk (sgolovatiuk) |
Changed in mos: | |
status: | Fix Committed → Confirmed |
Changed in mos: | |
importance: | High → Critical |
This is not a bug. If you are running keystone under apache, each apache restart will lead to keystone restart. If you want to avoid it, either run keystone as a separate process, or modify the environment so logrotate script is executed at random time.
If you still think this error is purely apache problem, please provide some arguments.