ZFS Pool Health Check Before Proxmox VE 9.2 Upgrade
Run a ZFS pool health check before upgrading Proxmox VE 9.2 to verify pool state, free space, and scans, preventing failed reboots and stalled backups.

On this page
Before you upgrade a Proxmox VE 9.2 node that uses ZFS, run a pool health check that proves every pool is ONLINE, has no errors, has enough free space, and is not busy importing, resilvering, or scrubbing. That one check turns a routine 9.1 to 9.2 update into a boring reboot instead of a storage rescue where backups, VM disks, and the boot menu all look guilty at once.
Key Takeaways
- Pool first: Check every ZFS pool before starting the Proxmox VE 9.2 upgrade.
- No busy pools: Never upgrade while a pool is degraded, faulted, resilvering, or scrubbing.
- Space matters: Confirm pool capacity and filesystem free space before snapshots, backups, or package upgrades.
- Boot matters: If ZFS is the root pool, verify the boot pool and mirror state before rebooting.
- Time it: A healthy check takes minutes; a failed upgrade can cost hours.
Why ZFS Health Changes the Upgrade Decision
A Proxmox VE 9.2 upgrade is not just a package bump on a ZFS host. The update can install a new kernel, refresh ZFS userland tools, and rewrite boot entries. If the pool is already carrying errors, the reboot is where that problem stops being background noise and becomes a failed boot, a stalled backup, or a storage pool that refuses to import cleanly.
ZFS is also doing more than one job on many Proxmox nodes. It may hold the operating system, VM disks, container root filesystems, and Proxmox Backup Server storage. A pool that is merely online in the GUI may still be a poor upgrade candidate if it is degraded, nearly full, or in the middle of a scrub.
For hosts that also run backup workloads, pool health ties directly to restore confidence. The same habits covered in the ZFS scrub schedule and ARC health for PBS 4.2 S3 backups article apply here: check the pool before you depend on it.
How to Run the 5-Minute ZFS Pool Check
Run these commands on each node that uses ZFS, preferably from the node shell with full root access. The commands are read-only and safe for a production check.
zpool status -x
zpool status -v
zpool list -o name,health,size,free,capacity
zfs list -o name,used,avail
The first command is the gate. If every pool is healthy, it returns a short clean message. If any pool has a problem, it prints enough to tell you that the upgrade should wait.
all pools are healthy
The second command gives the detail you need before making a decision. The last two commands show capacity, because a healthy pool that is nearly full can still cause upgrade-time surprises.
Check every pool, not just the root pool
Many Proxmox hosts have more than one ZFS pool. A default ZFS install may create one pool for the operating system, while a second pool holds VM disks, ISO images, or backup data. A third pool might exist for Proxmox Backup Server or an offsite sync target.
Check all pools, not only the pool that contains the root dataset. A data pool can be degraded while the root pool looks fine, and that still affects backups, snapshots, and storage operations during the upgrade.
If your host uses ZFS storage in the Proxmox GUI, the storage name may not match the pool name. The storage layer can look usable even when the underlying pool has lost redundancy or is short on space.
Check capacity before you snapshot or back up
Capacity is not a cosmetic detail. ZFS uses copy-on-write, snapshots, and metadata that need free space. A pool at 93 percent capacity may be online today and fail a snapshot tomorrow.
The pool capacity command gives a fast view of the pool itself:
NAME HEALTH SIZE FREE CAPACITY
rpool ONLINE 1.75T 1.31T 25%
tank ONLINE 12.0T 914G 93%
The filesystem list shows where space is being consumed:
NAME USED AVAIL
rpool 1.05T 2.4T
rpool/data 820G 2.4T
tank 11.2T 980G
For a homelab or small production node, I would not start an upgrade on a pool that is already near the edge. If snapshots, backups, or package upgrades need to write data, leave yourself headroom.
Check for background work that must finish first
A pool can be healthy and still busy. Scrubs and resilvers add read and write load. An upgrade reboot adds boot work, package unpacking, and storage access. If the pool is already trying to recover or verify itself, the upgrade window becomes less predictable.
The verbose status output includes a scan line. If that line says a scan is in progress, decide whether to wait or accept the extra load. For most single-node Proxmox hosts, waiting is the calmer choice.
What a Healthy Result Looks Like
A healthy upgrade candidate is more than a green icon. You want the pool state to be ONLINE, the error counters to be zero, and no active resilver or scrub. You also want enough free space for the upgrade and the first post-upgrade backup.
A healthy verbose result often looks close to this:
pool: rpool
state: ONLINE
scan: scrub repaired 0 in 01:22:14 with 0 errors on Sat May 10 03:14:00 2026
config:
NAME STATE READ WRITE CKSUM
rpool ONLINE 0 0 0
mirror-0 ONLINE 0 0 0
sda2 ONLINE 0 0 0
sdb2 ONLINE 0 0 0
errors: No known data errors
That output says the pool is online, the mirror is complete, the last scrub found no errors, and there is no scan running right now. That is the kind of result that lets you proceed without inventing a storage disaster scenario in your head.
Which Checks Actually Prevent Upgrade Surprises?
Boot pool surprises
If ZFS is the root pool, the upgrade touches the boot path. New kernel packages and bootloader entries can reveal a weak boot setup that was hidden while the node stayed online.
A mirror that is still online can still be a bad upgrade candidate if one member is missing, failing, or not present in the boot order. The node may boot fine from the remaining disk, but you have lost the safety margin you thought you had.
If you are moving from Proxmox VE 9.1 to 9.2, pair this pool check with the broader Proxmox VE 9.1 to 9.2 upgrade checklist. The pool check is the storage half of the decision.
Backup and storage surprises
Backups and snapshots are where ZFS capacity problems become obvious. A backup job can start happily, then fail when the pool cannot write the next block. Proxmox Backup Server may pause, a vzdump job may stop, or a snapshot may refuse to be created.
A Proxmox ZFS storage entry often points at a pool like this:
zfspool: local-zfs
pool rpool
content images,rootdir
sparse 1
If that pool is nearly full, the storage layer inherits the problem. The gotcha that caught me once was not a failed disk. The pool was online, but the root pool had climbed to 94 percent capacity after a week of test backups. The upgrade finished, the first snapshot after reboot failed, and PBS backups stalled until I deleted old snapshots.
That is why the capacity commands matter. A pool can pass the quick health check and still be a bad upgrade candidate because it has no room for the work the upgrade will ask it to do.
Performance and timing surprises
The honest tradeoff is time versus risk. Waiting for a scrub to finish can burn a maintenance window, especially on larger homelab pools. Upgrading anyway saves that window but can turn a hidden read error into a degraded mirror during the reboot and package update.
If the node is a single-disk ZFS pool, there is no redundancy to fall back on. A healthy pool is still a single point of failure. If it is a mirror or RAID-Z pool, a healthy check gives you a much better chance that the reboot will not expose a second failure.
A Simple Upgrade Gate for ZFS Nodes
Use the check as a gate, not as a suggestion. If the node passes, continue. If it fails, postpone the upgrade until the storage state is clean.
- Confirm that the quick health check returns a healthy result for every pool.
- Review the verbose status output for each pool.
- Confirm that pool capacity leaves enough headroom for snapshots and backups.
- Confirm that no resilver or scrub is running unless you accept the extra load.
- Run the package update and upgrade commands.
- Reboot during the planned maintenance window.
apt update
apt dist-upgrade
The gate keeps the decision simple. You are not guessing whether ZFS will behave during the upgrade. You are checking whether it already behaved before you changed anything.
| Pool state | Upgrade decision | Why |
|---|---|---|
| ONLINE, no errors, enough free space | Proceed | The pool is in a normal state for maintenance. |
| ONLINE, but scrubbing or resilvering | Wait | Added I/O can slow the upgrade and hide issues. |
| DEGRADED | Fix first | A reboot can turn lost redundancy into an outage. |
| FAULTED | Do not upgrade | Data availability is unknown until the pool is repaired. |
| ONLINE but nearly full | Clean up first | Snapshots and backups need headroom. |
What to Do When the Check Fails
Do not treat a failed check as an upgrade problem. Treat it as a storage problem that happens to be waiting for an upgrade to expose it.
- If a pool is degraded, identify the offline device, replace or reseat it, and wait for the resilver to finish.
- If a pool is faulted, stop and test the pool carefully before touching the Proxmox upgrade.
- If a pool is nearly full, delete old snapshots, prune PBS backups, or move images to another pool.
- If error counters are not zero, check the disks, cables, and SMART status before rebooting.
- If the pool is still importing, wait for the import to finish before you make the node more fragile.
If you need a wider backup-focused review, the ZFS pool health audit before PBS 4.2 backup jobs article covers the same idea from the backup side. The upgrade check is shorter, but the principle is the same: do not let a backup path sit on top of a questionable pool.
How Long Should You Wait Before Upgrading?
If the pool is healthy, there is no reason to wait beyond the normal upgrade window. If a scrub is running, waiting is usually the better choice. A 12 TiB mirrored pool can spend 6 to 10 hours in a scrub, and a failed disk resilver on the same pool can take another 2 to 4 hours depending on load.
That timing figure changes the decision. A 45-minute maintenance window is not a good match for a pool that may still be recovering. A weekend window is different. If the pool is healthy and the scrub finished an hour ago, you can usually proceed.
The tradeoff is still time versus risk. Waiting protects the upgrade window. Upgrading too early can turn a minor storage issue into a long recovery session after you have already changed packages and rebooted.
Conclusion
Checking ZFS pool health before upgrading Proxmox VE 9.2 keeps the upgrade focused on software changes instead of exposing hidden storage problems. Run the quick health check, the verbose check, and the capacity checks on every node with ZFS before you touch the upgrade. Your next step is simple: run the check now, record the result, and only upgrade nodes that pass the gate.


