[Labs-l] Switching controllers from virt1 to virt0, short downtime possible

Fri Jan 13 00:40:43 UTC 2012

I'm in the process of switching controllers from one virt1 to virt0.
I'm doing so so that we can move virt1 into the compute node pool,
increasing our storage, and increasing the number of instances we can
run.

virt1 is running a number of services that, if migrated improperly,
can cause downtime. I'll try to minimize this as much as possible.
Unfortunately, no matter what, the nova-api service will require
downtime during the transition, since I need to stop messages from
going into rabbitmq. This will only affect the ability to change
things in nova (like adding new instances, ip addresses, security
groups, etc).

If you notice any long running outage, please hop into #wikimedia-labs
and let me know.

- Ryan