Upgrading MAAS deployment

We have an all-in-one maas deployment still running on the test database (running in a VM). It is managing 120 servers in one IP space – all of the ipmi ports are in one class C subnet and all of the data connections are in another class B subnet. All of the VLANs are implemented at the switch ports so as far as MAAS is concerned, there are no VLANs. The vast majority of these servers are in use today and I can’t afford any interruptions to their availability.

Soon we we’ll be adding more servers in new racks in their own IP space. I think this means a second rack controller. And I need to move the server that hosts MAAS today.

Some of the things that I think need to change are

  • upgrade MAAS
  • upgrade to postgres
  • de-couple the rack controller from the region controller so we can have more than 1 rack controller.
  • straighten out the fabric/subnet config and migrate VLAN config to MAAS

Do I have these in a good order? Did I miss anything?

Is there any guidance on migrating from an all-in-one to dedicated rack controllers?

What are the implications if MAAS is down for an hour or two?

You have to strictly follow the instructions here. There is an upgrade path, which means if you directly upgrade MAAS from a lower version to a much higher version may cause inconsistency or failures.

  1. disconnect the subordinary controller (rack controller)
  2. perform a complete backup of the running database
  3. upgrade the PostgreSQL
  4. upgrade the master controller
  5. re-install the subordinary controller (Rack controller)
  6. adjust fabric configuration
1 Like