High Availability

High Availability (HA) ensures that critical VMs are automatically restarted on another node if the node they are running on fails.
Prerequisites:
A working cluster with at least 3 nodes
Shared storage accessible by all nodes (so the VM disk can be accessed after failover)
The HA manager service must be running on all nodes
Note: HA requires a quorum (majority of nodes must be online). With 3 nodes, you can tolerate 1 node failure. With 2 nodes, there is no quorum and HA will not function.
 
Enabling HA for a VM
Go to Datacenter > HA
Click Add under Resources
Select the VM ID you want to protect
Set the HA State:
    • Started — HA will always try to keep this VM running
    • Stopped — HA will manage the VM but leave it stopped
    • Disabled — HA does not manage this VM
Set Max Restart — how many times to try restarting on the same node before migrating
Set Max Relocate — how many nodes to try before giving up
Click Add
 
HA Groups
HA groups define which nodes are preferred or required for specific VMs.
Datacenter > HA > Groups > Add
Name the group and select nodes
Set priority (higher = preferred) for each node
Assign a VM to the group in HA Resources
 
Testing High Availability
Start an HA-protected VM on one node
Simulate a node failure by powering off that node or running:
systemctl stop pve-cluster corosync
Watch the Proxmox web console — the VM should automatically start on another node