ZFS mirror with hot spare

This lab walks through the most resilient ZFS topology used in Proxmox installations: a mirror vdev with a hot spare that activates automatically on disk failure.

Steps

1. Identify your disks

lsblk
# Identify three disks of equal size; for example /dev/sdb /dev/sdc /dev/sdd

2. Create the pool

zpool create datapool mirror /dev/sdb /dev/sdc spare /dev/sdd

Verify the topology:

zpool status datapool
# Expected output shows:
#   datapool
#     mirror-0
#       sdb
#       sdc
#     spares
#       sdd

3. Add Proxmox storage pointing at the pool

pvesm add zfspool datapool --pool datapool --content images,rootdir

4. Create a test VM on the pool

qm create 900 --name zfs-test --memory 1024 --cores 1 \
  --scsi0 datapool:32,iothread=1 --net0 virtio,bridge=vmbr0 \
  --ostype l26
qm start 900

Verify the VM is up and the pool reports no errors:

zpool status datapool

5. Simulate disk failure

Mark /dev/sdb as failed (replace with your actual mirror member):

# For a real disk: pull it. For a virtual disk in nested mode:
echo 1 > /sys/block/sdb/device/delete
# Or use zfs to mark it failed:
zpool offline datapool sdb

Watch the spare activate:

zpool status -v datapool
# Expected: sdd now active in mirror-0; sdb shows OFFLINE

6. Replace the failed disk

# Replace sdb with a new disk (physically or virtually)
zpool replace datapool sdb /dev/sde

# Watch the resilver
zpool status -v datapool
# Expected: resilver in progress, then completed; spare released

7. Verify final state

zpool status datapool
# Expected: mirror-0 with new sde, no spares active

Confirm the test VM is still running:

qm status 900

Verification

• The test VM survived the disk failure without interruption
• The spare activated automatically within seconds of failure
• After replacement, the new disk resilvered and the spare returned to standby
• No data loss occurred

Cleanup

qm stop 900 && qm destroy 900
zpool destroy datapool