Comment 4 for bug 1677434

I'm currently working with a 290 node Openstack cloud (2.1.2, Maas 2.2, Xenial & Ocata), and a 'juju status' takes at least 40 minutes.

I've mgopurged offline using v1.5, and have a cron running every 2 hours to prune.

We've restarted all 3 state servers multiple times, and there is no difference to the performance.

The state servers generally have a loadaverage >50, with 16 cores in each VM (3 controllers VMs).

Running 'juju status -m controller' takes between 3 and 60 seconds to run, depending on when the last restart was, but 'juju status' for the main model (running OpenStack) takes 40+ minutes regardless.

We're trying adding cores, and refreshing some of the controller VMs, but we really need more advice on how we can help the juju team diagnose this - what logs are helpful, etc.