On initial deployment non-leader units will show status message "Ports which should be open, but are not: 6642" until first update-status hook is run

[Frode Nordahl]

The OVN Northbound and Southbound OVSDB Server processes configure their listeners based on information in the database. On initial deployment the leader unit will set up the clustered databases and write the listener configuration to it.

Subsequently the non-leader units will join the cluster in the leader-settings-changed hook, but often the data will not be synchronized and listeners configured by the time the first call to assess_status.

While this condition is benign, it may cause confusion and prolonged deployment time.

Perhaps we could detect the condition and have the charm stick around for a bit and retry the assess status a few times/until it's green?