Created attachment 1148443 [details]
the ansilbe logs for pre.yaml

Situation here should be improved quite a bit today, afaik there is nothing in pre.yml that should take down any services so these should not affect running systems.

Once we proceed to upgrade itself we upgrade each master 1 at a time, then for every node (which includes masters) we set it unschedulable, then upgrade docker if required, then upgrade the node/openvswitch and restart services. I believe service downtime is about as minimal as possible at this point, I can't see any extraneous tasks in the plays between taking the service down or node offline, and when they come back.

Sending back to you Anping, let me know if you think there is anything else that can be done. If not lets close this.

I works fine now. So move to verified

Moving this to private as I don't think there's any relevant doctext to be written here.