Attempting the big sliding tile puzzle: moving customers to new virtualization hosts so that I can empty out one machine and upgrade its kernel, then move things back, so I can rinse and repeat twice more.
Replies:
6
@insom i don't mind several hours of downtime today for my vms if that makes any of that process easier
@pho4cexa Thanks! But the goal is to be _better_ than big providers though, and not just cheaper, so we're going for zero-downtime here.
I just migrated our first machine from DRBD-backed storage and into Ceph without even dropping a ping packet, and that's the plan for the rest of the machines.
Everyone was migrated away from castor without noticing too 鈽猴笍 (although exactly one machine caught a stray because of its running config not matching its stored config on disk, regrettably).
