I’m planning to use OpenNebula for KVM deployments (production usage).
I’m mostly concerned with storage types available.
A. OpenNebula controller (master server where only the panel will be hosted)
Does it need to be hosted on the same network as the hyperivsors/virtualisation hosts or somewhere off-site (like a full virt. virtual machine) would work? This will reply on public interface for communication to the virt. hosts.
B. Virtualisation hosts:
I am planning to have two virtualisation hosts of the same configuration - with hardware RAID 10 and HDD/SSDs. I will rely completely on local storage and no SAN or other network-based storage. I will also use LVM and ensure the creation of a empty volume group for the RAW VM disks to be mounted.
How would I setup local storage utilising LVM on OpenNebula?
Should I also setup a private network on my switch, with each virtualisation host connected on 1GE links? (in cases for migrations through SSH copy etc)
Hm. I think I don’t know exactly where your questions aim to.
Why do you want Opennebula running on a public IP?
If you specify this, maybe I can give you a better suited answer, but generally:
If you have shared storage, both Opennebula frontend and the Opennebula nodes (KVM hosts) will need to have access to it (e.g.: NFS or Ceph), if you use local storage for the running images (system datastore) you still need access from Frontend to image datastore and then as well need to have some kind of network connection from frontend to nodes that can handle the copying of the images.
If you only want to know if you can have sunstone (the webinterface itself) listening on a public IP: yes, why not, but I would have some kind of SSL offloading in between.