IT 3300 : Virtualization

Proxmox — Cluster & High Availability

Why cluster?

• Manage all your nodes from one web UI
• Move workloads between nodes
• Enable high availability: auto-restart VMs when a node dies

Building a cluster

• On the first node: Datacenter -> Cluster -> Create Cluster
• Copy the join information
• On each other node: Join Cluster, paste the info
• All nodes now appear under one datacenter

Quorum

• A cluster makes decisions by majority vote (quorum)
• Prevents "split brain" where two halves both think they're in charge
• Need more than half the votes to act
• This is why an odd number of nodes (3, 5) is preferred

Migration

• Offline migration — move a stopped VM to another node
• Live migration — move a running VM with no downtime
  • requires shared storage (NFS/Ceph) so the disk is reachable
  • memory state is copied over, then execution hands off

High availability (HA)

• Mark a VM as HA-managed
• If its node fails, the cluster restarts it on a healthy node
• Requires shared/distributed storage and quorum
• Recovery is automatic, not instant — the VM reboots elsewhere

Fencing

• Before restarting a VM elsewhere, the cluster must be sure the old
  node is truly gone (to avoid running it twice)
• Fencing isolates the failed node
• Quorum loss triggers protective shutdown of the minority side

Putting Unit 1 together

You now have a real virtualization platform:

• Multiple nodes in a cluster
• Shared storage
• VMs that survive a node failure

This is the foundation the rest of the course runs on.

Lab goals

• Join your nodes into one cluster
• Live-migrate a running VM between nodes
• Mark a VM HA, pull the plug on its node, watch it recover