Hello, you should use ha-groups for that and pin vms to different hosts.
For example: vm1 pinned to host1/2 and vm2 pinned to host 3. or different, depending on what cluster-node-count you have.
Hi @skwal , to answer your question - there is no "one-checkmark" built into PVE to achieve your goal.
Given a sufficient number of hosts in the cluster, the solution proposed by @jsterr can work.
However, if the number of nodes is small - then it becomes increasingly harder.
Realistically, with 5 nodes you could:
VM1 - restrict to n1,n2
VM2 - restrict to n3,n4
Manually add/remove n5 on the failure of any one host in group 1 or 2.
You could do 4+qdevice.
With 3 nodes you will need a procedure in place to recover manually after failure.
Sadly, the more reliable approach is an outside script that polls the cluster often and moves VMs around as needed.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.