Pause action doesn't stop all services

Bug #2003850 reported by Przemyslaw Lal
14
This bug affects 3 people
Affects Status Importance Assigned to Milestone
OpenStack Designate Charm
In Progress
Undecided
Unassigned

Bug Description

Running the "pause" action on designate units doesn't stop apache2 that listens on port 8991 which is used by haproxy to determine whether the service is up. As a result, haproxy on other units still ends up forwarding requests to the "paused" unit which results in errors and API unavailability, e.g. in the CLI client:
$ openstack zone list
Failed to contact the endpoint at https://designate.public.example.openstack:9001 for discovery. Fallback to using that endpoint as the base url.
Unknown

Steps to reproduce:
1. 3 designate units up and running, related to hacluster charm.
2. Pause one unit.
3. Send requests to designate VIP on port 9001 and observe haproxy.log on the unit with active VIP. It still tries to forward requests to the paused unit.

Workaround:
SSH into the paused unit and stop apache2:
$ sudo systemctl stop apache2

After that, haproxy on the remaining active units stops forwarding requests to the paused unit and the functionality is restored.

Channel: latest/stable
Revision: 60

Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix proposed to charm-designate (master)
Changed in charm-designate:
status: New → In Progress
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.