Proxmox VE: ZFS, LVM-thin, qcow2. Recovery across all three storage stacks.
Proxmox VE is the open-source virtualisation platform we see most often in our cases. Typical storage layouts: ZFS (pools, datasets), LVM-thin (thin-provisioned logical volumes), qcow2 (QEMU image files). Our recovery work covers all three layers and the ways they interact.
Faulted pools, orphaned snapshots.
ZFS is an integrated filesystem-RAID with strong self-healing capabilities. When a pool enters the faulted state it can often be recovered with zpool import -fF in read-only mode, where necessary with a rewind to an earlier checkpoint. In more severe scenarios we clone the disks and algebraically reconstruct the missing vdevs.
Thin pool corrupted, metadata exhaustion.
LVM-thin is space-efficient (on-demand allocation) but has a well-known weak point: if the metadata area fills up or becomes corrupted, the entire thin pool turns inaccessible. Recovery requires working directly on the thin pool metadata and rebuilding the logical volume manager on a cloned volume.
Corrupted QEMU images, snapshot chains.
qcow2 (QEMU Copy-On-Write) is the standard format for Proxmox VMs when storage is neither ZFS nor LVM. Header, L1/L2 tables, multiple internal snapshots, backing file chains: every one of these can become corrupted and each requires dedicated recovery. Tooling: qemu-img check, rebuilding damaged L1/L2 tables, and manual extraction of raw data when the qcow2 cannot be recovered as such.
The questions we get asked most.
Proxmox VE cluster with 3+ nodes: how does recovery work if one node fails?
If HA replication is enabled and correctly configured, critical VMs should fail over automatically. Recovering the failed node is a separate job: usually hardware recovery of the server, then rejoining the cluster. Split-brain or lost quorum requires careful manual intervention on the quorum tool (pmxcfs).
Is it worth migrating to ZFS or should we stay on LVM-thin?
ZFS brings significant advantages (compression, atomic snapshots, end-to-end checksums, scrub) but demands more RAM (ARC). On systems with 64+ GB of RAM dedicated to storage, ZFS is almost always the better choice for new deployments. On smaller or constrained systems, LVM-thin remains a valid option.