Unable to use smartctl for SCSI Disks in OMV VM on Proxmox

fonzdm

Hi everyone!
I'm facing an issue in reading SMART data for a bunch of SCSI Hard Drives from OMV.
OpenMediaVault is running as a dedicated VM on top of my Proxmox VE.
I have three disks (one standalone, two in raid) that are directly mapped to my OMV VM:

{disk list on Proxmox?}

In OMV, Storage → S.M.A.R.T. → Devices

{disk list on OMV?}

As you can see, the temperature is not shown correctly (0°C) and the Status is Unknown.
If I manually run a short smart test, I get:

=== START OF INFORMATION SECTION ===
Vendor:               QEMU
Product:              QEMU HARDDISK
Revision:             2.5+
Compliance:           SPC-3
User Capacity:        4,000,787,030,016 bytes [4.00 TB]
Logical block size:   512 bytes
LU is thin provisioned, LBPRZ=0
Device type:          disk
Local Time is:        Tue Sep 16 16:37:52 2025 CEST
SMART support is:     Unavailable - device lacks SMART capability.
Read Cache is:        Enabled
Writeback Cache is:   Enabled

=== START OF READ SMART DATA SECTION ===
Current Drive Temperature:     0 C
Drive Trip Temperature:        0 C

Error Counter logging not supported

Device does not support Self Test logging
Device does not support Background scan results logging
Short offline self test failed [unsupported scsi opcode]

END OF LINE

While everything is fine if performed directly from Proxmox, for example:

{smart data from Proxmox?}

Do you how can I solve this?
I have set up alerts on OMV and I would like it to monitor the HDDs, so in case of excessive temperature and/or RAID corruption (managed directly by OMV), I get an email.

Thanks in advance,
fonzdm / a human

imgios

Hi fonzdm, I'm not 100% sure how Proxmox works, but I can see that the official way to achieve that is using PCIe Passthrough.

For sure you already know that, and if you are looking for a dirty solution (😉) I think you can replace the smartctl (create a copy) binary with a bash script. I wrote this bash script that does the following:

Parse all the arguments passed by OMV:
- The disk argument is read and replaced by the target if it's a symlink
Based on the disk it will run the smartctl command on the OMV VM if it's the root disk, otherwise will connect to the Proxmox host (where you have your disks) and run the smartctl there. This only works if the disk files attached in the OMV VM are the same in the Proxmox host.

Before running the script, make sure to (1) rename the smartctl bin in your OMV VM to orig.smartctl, (2) replace the placeholders with your root disk and the Proxmox host user/IP. I suggest to use ssh-copy-id to let the connection happen via key.

#!/bin/bash

new_args=()

fetch-smart() {
	islocal=$1
	shift
	if [[ "$islocal" == "true" ]] ; then
		orig.smartctl $@
	else
		ssh -q <username>@<proxmox-host> "smartctl $@"
	fi
}

localRun="false"

for arg in "$@"; do
    # Only consider arguments starting with /dev/
    if [[ "$arg" == /dev/* ]]; then
        # Replace it with the resolved path
        resolved=$(readlink -f "$arg")
        new_args+=("$resolved")
        if [[ "$resolved" == "/dev/sdd" ]] ; then
			localRun="true"
        fi
    else
        # Keep other arguments as-is
        new_args+=("$arg")
    fi
done

fetch-smart "$localRun" "${new_args[@]}"

Save the script as smartctl, make it executable with chmod +x smartctl and place it in the same folder where the original smartctl is stored (you can find it out using which smartctl before running al these steps).

This could probably be improved by using disk UUIDs, but I'm not sure if they match the host's UUIDs.

fonzdm

HOLY COW 🧐

I like this dirty little script! Yes, I knew about the PCIe option but I like the script better: it allows me to fetch SMART data and perform smartctl tests without having to mess with the Proxmox config (again).
I did some tweaks to the script: your version is rock solid, but suffer one major problem

sda, sdb devices (and so on) are not immutable ⚠️

So what happens at host or guest reboot? If the device symlinks do not match between host and guest, we end up fetching SMART data from the wrong disk!
One major drawback for this is the smartctl test (short or long, doesn't matter): you may end up testing the wrong disk! This could also impact the RAID status and so on.
What I did was integrate a strict mapping between guest-level disks-by-id and host-level disks-by-id: using a map file, I can be 100% sure that the disks OMV thinks to be testing is its real counterpart at host level.
At the moment, I'm using a static mapping file: dynamic solution are welcome, as long as they do not over-load, over-complicate the setup.
Here's the script:

#!/bin/bash

MAP_FILE="/opt/host-guest-disk.map"
new_args=()

fetch-smart() {
    islocal=$1
    shift
    if [[ "$islocal" == "1" ]] ; then
        orig.smartctl "$@"
    else
        ssh -q <proxmox user>@<proxmox-hostname/ip> "smartctl $@"
    fi
}

# Default: run locally unless the map says otherwise
localRun="1"

get_disk_byid() {
    dev="$1"
    # Get ID_SERIAL from udevadm
    serial=$(udevadm info --query=property --name="$dev" | awk -F= '/^ID_SERIAL=/{print $2}')
    bus=$(udevadm info --query=property --name="$dev" | awk -F= '/^ID_BUS=/{print $2}')
    type=$(udevadm info --query=property --name="$dev" | awk -F= '/^ID_TYPE=/{print $2}')

    # Construct expected by-id path
    if [[ -n "$serial" && -n "$bus" ]]; then
        echo "/dev/disk/by-id/${bus}-${serial}"
    else
        # Fallback: just return the original device
        echo "$dev"
    fi
}

for arg in "$@"; do
    if [[ "$arg" == /dev/* ]]; then
        # If user gave plain /dev/sdX, convert to /dev/disk/by-id/...
        if [[ "$(basename "$arg")" =~ ^sd[a-z]+$ ]]; then
            byid=$(get_disk_byid "$arg")
            short_id=$(basename "$byid")
        else
            short_id=$(basename "$arg")
        fi

        # Look up mapping
        line=$(grep -m1 -F "$short_id" "$MAP_FILE")
        if [[ -n "$line" ]]; then
            local_id=$(echo "$line" | cut -d: -f1)
            is_local=$(echo "$line" | cut -d: -f2)
            remote_id=$(echo "$line" | cut -d: -f3)

            new_args+=("/dev/disk/by-id/$remote_id")
            localRun="$is_local"
        else
            # No mapping → keep original and force local run
            new_args+=("$arg")
            localRun="1"
        fi
    else
        new_args+=("$arg")
    fi
done

fetch-smart "$localRun" "${new_args[@]}"

An example of the mapping file:

scsi-0QEMU_QEMU_HARDDISK_drive-scsi1:0:ata-ST000000000-XXXXX_XXXXXX
scsi-0QEMU_QEMU_HARDDISK_drive-scsi2:0:ata-ST0000V0000-YYYYY_YYYYYYY
scsi-0QEMU_QEMU_HARDDISK_drive-scsi3:0:ata-ST000000000-ZZZZZ_ZZZZZZZ
ata-QEMU_HARDDISK_QM00001:0:nvme-SPCC_M.2_PCIe_SSD_000000_XXXXXX
ata-QEMU_DVD-ROM_QM00003:1:ata-QEMU_DVD-ROM_QM00000

The structure is really simple

<guest-level-disk-id>:<is-local>:<proxmox-level-disk-id>

The middle parameter, <is-local>, is just used to decide if the smartctl command needs to be run locally or remotely.

Overall, this seems to work! 🎉

Thanks a lot for your help!
@fonzdm / a human

imgios

I love it! These are nice improvements — this is what I was talking about at the end of my previous post. I'm happy that this dirty solution solved your OMV "limitation" 😊

Please let me know if you would prefer to lock the discussion, or if you would like to keep it open in case anyone else has a better solution.

lucaam

Uanm, complimentoni