Our Proxmox cluster relied on NFS storage approaching end-of-life. We planned a live migration to a Ceph-backed pool without VM downtime, using Proxmox live storage migration and staged workload groups.
Each migration window included pre-flight IOPS benchmarks, Ceph health checks, and rollback snapshots. Network isolation between Ceph public and cluster networks was validated under production load before cutover.
The migration completed over three weeks with zero downtime. Performance improved 30% on random I/O workloads. The main lesson: never migrate storage and hypervisor firmware in the same window.