← All break/fix scenarios in Proxmox VE
The switch port for the corosync VLAN was not brought back up after maintenance. Corosync on ring 1 (the dedicated cluster link) is partitioned, so each node sees itself as the only member of a single-node partition and refuses to take cluster actions. A common alternative cause is asymmetric routing on the corosync VLAN — one node still has its old default route for that subnet.
1. Verify the switch port: `show interface status` on the switch for the corosync ports. 2. Bring the port up: `no shutdown` on the corresponding interface. 3. On each node, bounce the corosync link: `ifdown <corosync-iface>; ifup <corosync-iface>`. 4. Recheck: `pvecm status` should return to Quorate: Yes. If the link is up but corosync still won't converge, check that the VLAN is correctly trunked through any intermediate switches.
- `pvecm status` shows Quorate: Yes with 3 active votes - `pvecm nodes` lists all three nodes - `ha-manager status` returns without errors - VMs can be started, stopped, and migrated normally
- Run the corosync link on a physically separate switch if possible - Document the maintenance procedure to include port verification - Consider deploying a QDevice as an additional tie-breaker - Add a corosync health check to the daily checklist
Read the symptoms and evidence. Form a hypothesis. Then reveal the diagnosis and remediation to compare your reasoning.
The fix in this case is operational (network) rather than configuration — that’s typical for corosync issues. Check the physical layer first, software layer second.