openNebula is a very cool project!
Please be gentle, I’m new here
I want to build the bones of an OpenNebula cloud with five servers (2 x datastores & 3 x KVM hosts), I have a few questions though:
Should I use identical hardware for all servers (datastores and hosts)?
Should I virtualise the datastore servers (probably will be Ubuntu x86_64, ZFS and GlusterFS)?
Can I install the OpenNebula servers (in HA) on the KVM hosts, the datastore servers or separate servers altogether?
I’m kind of thinking it might be cool to have identical servers that can take on either role but it would be costly in terms of all the servers having to have host NIC’s, datastore HBA’s and disks installed to cover any role?
So maybe I have two hardware builds, a KVM host server and a datastore server?
Also could I collapse my datastore and KVM hosts in to a single server, the host OS handling the ZFS (can OpenNebula still manage this?), my (qcow2) server images and GlusterFS running in KVM VM’s?
I don’t need very high disk performance or space requirements so I was thinking of starting off with each server having one 800GB enterprise SSD running ZFS with dedup, compression and “copies=2” and having GlusterFS also making two copies?
I have just realised there are a lot of different questions about a lot of different technologies, sorry about that.
Thanks for your time.