Comment 19 for bug 1846789

I'm working on rocky release and have the same problem. It happens on each deploy on three-node cluster. I haven't notice this problem on single-node deployment.
After deploying keystone, its endpoint (port 5000) is quite unstable. It works five, six times and then fails with "500 Internal Server Error". Then it works again some time and again fails. When Internal server error is returned, the keystone error logs are the same as Larry had:

mod_wsgi (pid=20): Target WSGI script '/var/lib/kolla/venv/bin/keystone-wsgi-admin' cannot be loaded as Python module.
mod_wsgi (pid=20): Exception occurred processing WSGI script '/var/lib/kolla/venv/bin/keystone-wsgi-admin'.

I checked on keystone container on each node and there is no rule, sometimes only one has this problem, sometimes two.
Cloud deploymend then fails when some other module needs keystone endpoint.

I found out, that restarting keystone containers and redeploying cloud resolve this situation.

One more note, I did't have those problems with the rocky version from Nov 7th and I and my collegues deployed single and multicloud many times.