Restore unavailable Proxmox infrastructure
Impact
A Proxmox host, datastore, bridge, or group of Talos VMs is unavailable. The failure can remove several Kubernetes and etcd members in one failure domain.
Triage
Use the independent Proxmox management path to check host quorum, VM power, ssd-zfs, free space, pool health, vmbr20, recent migration and backup jobs, and hardware events. Do not start duplicate copies of the same VM disk or identity.
Immediately check surviving Talos and etcd members from another failure domain:
Decide what to do
- If one VM failed while Proxmox and storage are healthy, use Talos node unavailable.
- If one control plane failed and etcd retains quorum, use node replacement.
- If multiple control planes failed or the datastore is lost, preserve surviving disks and backups. Then use disaster recovery.
- If Proxmox cluster quorum is at risk, stabilize Proxmox before changing Talos.
Check the resolution
Resolve the alert only when Proxmox quorum, storage, networking, and all intended VMs are healthy. etcd must have a leader and the expected membership. Kubernetes nodes must be Ready. Affected workloads and backup jobs must have recovered.