HAProxy marks swift backends as down because xinetd didn't start swiftcheck service
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Fuel for OpenStack |
Fix Released
|
Critical
|
Artem Panchenko | ||
6.0.x |
Invalid
|
Undecided
|
Unassigned |
Bug Description
After successful deployment BVT failed while checking HAProxy back-ends:
AssertionError: HAProxy back-ends are DOWN. {'exit_code': 0, 'stderr': [], 'stdout': ['swift node-4 Status: DOWN/L4CON Sessions: 0 Rate: 0 \n', 'swift node-5 Status: DOWN/L4CON Sessions: 0 Rate: 0 \n']}
Swift back-end was marked as DOWN on non-primary controllers, because 'swiftcheck' service didn't run on it. According to puppet logs 'xinetd' service was reloaded before 'swiftcheck 49001/tcp' line was added to /etc/services file (node-4):
http://
<27>Jun 4 19:29:00 node-4 xinetd[10557]: service/protocol combination not in /etc/services: swiftcheck/tcp
Steps to reproduce:
1. Create cluster: CentOS, NovaVlan
2. Add 3 nodes with controller roles
3. Add 2 nodes with compute roles
4. Set up cluster to use Network VLAN manager with 8 networks
5. Deploy the cluster
6. Check HAProxy back ends
Expected result:
- all back-ends are UP
Actual result:
- back-ends for Swift are DOWN on some controllers
Fuel version info (6.1 build #507): http://
Diagnostic snapshot is attached.
Changed in fuel: | |
importance: | High → Critical |
tags: | added: staging |
Fix proposed to branch: master /review. openstack. org/188621
Review: https:/