How to move a self-hosted OpenNebula to another machine

I’m currently running OpenNebula Community Edition 6.2.0 on an old server (centos 7. I’ve planned to migrate my centos installations to almaLinux, but this server can’t be upgraded to some of the hardware).

My objective is to move, or install a new OpenNebula, while my vms are still running on the hosts. I’m not sure if there is a “built in” solution to do that ? I’ve been pondering a little about in what direction I should look. Here are my main ideas :

  • Temporarily upgrade to an HA installation, registering my new machine as a slave. I imagine that I can kind of sync them that was, then use my new machine as master and shutdown the old one.
    (I’m not yet sure if it’s possible or possibly just a bad idea)

  • Potentially make an installation from scratch (Not sure yet if I should install a 6.2.0 or a more recent one) and somehow transfer all the configuration ?

Maybe just try to clone the current machine and restore it on the new machine ? But seems the hardware could be pretty different, I presume I’ll get some more os related issues.

Well if someone has already done that kind of migration, or if you have any advice. I’ll be glade to have another point of view on that situation.

I’m using a kvm hypervisore. I have 4 hosts, 2 Vnet, and 3 datastores (ceph managed)

Essentially, the OpenNebula frontend is

  • the OpenNebula software, which is usually packaged nicely for easy of setup and maintenance
  • the configuration for it
  • a database that holds the state

If you want to move to another host, you can simply install the same packages and import the database and configuration from the old instance. This is essentially what happens during an upgrade, but within the same host.

hey, reading about your circumstances, you already have shared storage for your VM disks, so I would add your new hosts as ONE hypervisors. Then you should be able to live migrate all your VMs to those hosts and keep them running.
Even if you kill or reinstall or update the ONE master server, the VMs should be OK.
Like Daniel said, if you want to move to a new master server, you can make a DB dump using the onedb command, save your configuration-changes from the /var/lib/one directory (oned.conf)
and migrate to a new master-server.
Running VM’s and their details should migrate with your DB dump, as long as all hosts can communicate with Ceph, this should work without interruption.
While moving all this functionality, I would upgrade to the next LTS version (6.8, right?) when that comes out somewhere this year, and you’ve upgraded all your components to the latest version.

If you install a new master server, import the DB and connect it to the datastores, you should be able to migrate without any downtime for your VMs

I’ve planned to upgrade from the 6.2.0 to the next LTS yes. But I presume I should first install the same version I currently have to import the DB and the configurations files.

If you upgrade, a copy of the current DB is saved and a copy of all configs in /etc/one as well,
so either way, you would have a way back.