ZFS interrupted resilver
ZFS resilvering reconstructs allocated blocks onto a replacement or returning member. If another device fails or reports read errors mid-process, the pool may become degraded, suspended, or unable to complete the intended redundancy transition.
Understand the symptom
ZFS uses checksums, transaction groups, and vdev redundancy to select valid blocks. At interruption, some allocated blocks may have reached the new member while others still depend on the old set. Recovery must preserve every participant and reconstruct the correct topology before choosing an import or extraction path.
What you may see
A source or destination device cannot complete required reads or writes.
ZFS stops I/O after storage failures exceed safe continuation.
Only allocated ranges processed before interruption contain reconstructed data.
Current hardware paths or labels no longer match the saved vdev topology.
Protect the source
An old or partial member can still contain blocks absent from the replacement.
Mounting writable can change labels, metadata, and pool state.
Each attempt rereads broad allocated regions from marginal disks.
Do this now
Pause shares, VMs, databases, scrubs, resilvers, and automatic replacement activity.
Save zpool status, events, history, labels, GUIDs, topology, errors, and import output.
Record original slot, device path, serial, removal time, replacement role, and resilver progress.
Preserve readable regions before another import or resilver attempt.
NEPFOSYS process
NEPFOSYS maps vdevs, GUIDs, labels, transaction state, and original versus replacement members.
Unstable and partial members are protected independently.
Candidate states are validated with checksums, metadata, and requested datasets before file recovery.
A clear next step
Start a case with the exact symptoms, incident timeline, prior attempts, encryption details, and the files that matter most. NEPFOSYS will route the case to the appropriate DFW intake or nationwide mail-in path.
Start with a case reviewQuestions, answered
Yes. It may hold valid reconstructed blocks, so preserve it with the old member and all survivors.
Not on the only copy. A force option can select an older or risky transaction state and may permit new writes.
Checksums identify corruption; repair still requires another valid copy from mirror, parity, snapshot, backup, or another source.
DFW local · Nationwide mail-in