HTTPS request to designate randomly fail

[David Negreira]

The following happens when we have a juju designate cluster with HA and SSL enabled.

When a unit is put on pause, for example with:

- juju run-action designate/2 pause

The apache service is not stopped, and the haproxy from pacemaker/corosync will still route the connection and the server will return 503, since the designate service is stopped but not apache.
The connections to the other units that are not paused work well as expected, but in a cluster of three servers we will get a 503 every three requests.

The fix should be to add the apache2 to the list of services and stop it when necessary.

[Dr. Jens Harbott (j-harbott)]

Seems this is not a bug of designate, but of the juju deployment.

[David Negreira (dnegreira)]

Hi @j-harbott

I am not sure how this is an issue of the juju deployment, if you pause the units it should stop the apache2 service and no connections should be made there.
This is currently not happening on any deployment.
I have pushed a fix to try to solve this: https://review.opendev.org/c/openstack/charms.openstack/+/920474/1

[David Negreira (dnegreira)]

Sorry, I just now saw that I have open a bug for -designate- and not the -designate-charm-, my apologies.