“OpenNebula can be (and it is usually) installed in a virtual machine.
Therefore the previous one-server1 and one-server2 can be in fact
virtual machines running in the same physical host (you can run them
in different hosts, requiring a different fencing plugin)”
I am wondering what is the best practice for running Opennebula as VM. Can anyone comment on that ?
Do we use virt-install to create VMs ? Virt manager ? Since KVM is installed this should not be a problem.
Any comments ?
We have two frontends and two sunstone servers in seperate VM’s on seperate hosts. We manage them through virsh / virt-manager. We have one primary master (running opennebula) and one slave that use MySQL master-master replication. We don’t do automatic failovers (not for opennebula, not for vm’s running on hosts) so no fancing is needed.
Thank you . I remember you from Berlin Conference and we had good time ! Do you remember ? I would like to know why you chose to run opennebula as VM when you had physical hosts at your disposal .
I am wondering what is the best practice for running Opennebula as VM. Can anyone comment on that ?
Do we use virt-install to create VMs ? Virt manager ? Since KVM is installed this should not be a problem.
Any comments ?
Hello,
We use corosync/pacemaker with a shared /var/lib/one/datastores.
We actually do not have redundant setup, the frontend VM is started
somewhere else when a node get fenced.
Here is the cleaned pacemaker configuration if you want to see how it
looks: pcmk.conf (7.56 KB)