Haproxy resource doesn't check nonlocal_bind option for already created namespace
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Fuel for OpenStack |
Fix Committed
|
Medium
|
slava valyavskiy | ||
Mitaka |
Fix Released
|
Medium
|
slava valyavskiy | ||
Newton |
Fix Committed
|
Medium
|
Oleksiy Molchanov | ||
Ocata |
Fix Committed
|
Medium
|
slava valyavskiy |
Bug Description
Detailed bug description:
I encountered a case when 'ip_nonlocal_bind' option was set to '0' during the cluster re-deployment.
"""
root@node-1:~# ip netns exec haproxy /sbin/sysctl net.ipv4.
0
"""
As result, haproxy resource wouldn't able to start.
I didn't find an enemy service what had changed this value, but, anyway, I suppose it would be nice to add corresponding check into ocf script.
Steps to reproduce:
Re-deploy fuel with changed number of database or any other nodes what service is under Haproxy control.
Expected results:
Haproxy resource works fine.
Actual result:
Haproxy resource is broken.
Reproducibility:
One time in several re-deployments.
Workaround:
No
Impact:
Re-Deployment is failed
Description of the environment:
Versions of components: MOS-10.0
Additional information:
Not applicable
Changed in fuel: | |
milestone: | none → 10.0 |
tags: | added: on-verification |
Fix proposed to branch: master /review. openstack. org/384982
Review: https:/