Manual Pure vVOL VMware to Proxmox Migration Guide

1.1 Get VM Information from vCenter

1. Log into vCenter

2. Find the VM and note:

   • VM Name

   • Number of virtual disks

   • Datastore (should be vVol-based)

   • MAC addresses (for network preservation)

   • Firmware type (BIOS or EFI)

1.2 Identify vVol Volume Names on Everpure

Using Everpure GUI:

Using purecli:

# List all volumes
purevol list

# Filter by VM name pattern
purevol list | grep -i "<vm-name>"

# Get volume details including serial
purevol list --csv | grep -i "<vm-name>"

2.1 Graceful Shutdown via vCenter

1. Right-click the VM in vCenter

2. Select Power > Shut Down Guest OS

3. Wait for the VM to fully power off

2.2 Alternative: Force Power Off (if guest tools not responding)

1. Right-click the VM

2. Select Power > Power Off

3.1 Using Everpure GUI

3.2 Using purecli

# Clone each vVol to a new volume
purevol copy <source-vvol-name> <new-volume-name>

# Example:
purevol copy "vvol-myvm-abc123-virtualDisk-0" "proxmox-myvm-disk-0"

# Get the serial of the new volume
purevol list proxmox-myvm-disk-0 --csv

# Repeat for each data volume

3.3 Record Clone Information

For each cloned volume, record:

• Volume name: proxmox-myvm-disk-0

• Serial number: (from purevol list)

• WWN: 3624a9370 + serial_lowercase

• Size in GB

Before attaching the cloned LUNs to Proxmox, create filters to prevent LVM from activating volume groups and kpartx from creating partition mappings on the cloned disks.

4.1 Add WWN to Filter List

We will use /var/lib/migrator/filtered_wwns to track all WWNs currently being migrated. This serves several purposes:

• Concurrent migrations: Multiple VMs can migrate at once, each with their own disks

• Safe cleanup: When one migration finishes, only that disk's filter is removed while others remain protected

• Persistence: The list survives reboots if a migration is interrupted

On ALL Proxmox nodes, run:

# Create directories
mkdir -p /etc/lvm/lvm.conf.d
mkdir -p /var/lib/migrator

# Add WWN to the filter list (repeat for each disk)
# This file tracks all active migration LUNs for safe cleanup later
# Replace the example WWN with your actual WWN from Step 3
echo "3624a93708eabcb40cc4241b208501082" >> /var/lib/migrator/filtered_wwns

4.2 Create LVM Filter Configuration

Create /etc/lvm/lvm.conf.d/pure-migrator.conf:

# Replace the example WWN with your actual WWN from Step 3
cat > /etc/lvm/lvm.conf.d/pure-migrator.conf << 'EOF'
# Filters migration LUNs to prevent LVM activation
devices {
    global_filter = [ "r|/dev/mapper/3624a93708eabcb40cc4241b208501082|", "a|.*|" ]
}
EOF

For multiple disks, add each WWN as a reject rule:

global_filter = [ "r|/dev/mapper/WWN1|", "r|/dev/mapper/WWN2|", "a|.*|" ]

4.3 Create Udev Rule to Block kpartx

Create /etc/udev/rules.d/59-migrator-skip-kpartx.rules:

# Replace the example WWN with your actual WWN from Step 3
cat > /etc/udev/rules.d/59-migrator-skip-kpartx.rules << 'EOF'
# Skip kpartx for migration LUNs to prevent partition mapping
KERNEL!="dm-*", GOTO="migrator_end"
ACTION!="add|change", GOTO="migrator_end"
ENV{DM_UUID}!="mpath-?*", GOTO="migrator_end"
ENV{DM_NAME}=="3624a93708eabcb40cc4241b208501082", ENV{DM_SUBSYSTEM_UDEV_FLAG1}="1", GOTO="migrator_end"
LABEL="migrator_end"
EOF

For multiple disks, add a line for each WWN before the LABEL line:

ENV{DM_NAME}=="<wwn1>", ENV{DM_SUBSYSTEM_UDEV_FLAG1}="1", GOTO="migrator_end"
ENV{DM_NAME}=="<wwn2>", ENV{DM_SUBSYSTEM_UDEV_FLAG1}="1", GOTO="migrator_end"

4.4 Reload Udev Rules

On ALL Proxmox nodes:

udevadm control --reload-rules

5.1 Using Everpure GUI

For each cloned volume:

5.2 Using purecli

# Connect volume to host group
purevol connect <volume-name> --hgroup <host-group-name>

# Example:
purevol connect proxmox-myvm-disk-0 --hgroup proxmox-cluster

# Verify connection
purevol listconnection proxmox-myvm-disk-0

On ALL Proxmox nodes, run:

# Using rescan-scsi-bus.sh (recommended)
rescan-scsi-bus.sh -a -r -u

# If rescan-scsi-bus.sh not found, install sg3-utils first:
apt-get install -y sg3-utils
rescan-scsi-bus.sh -a -r -u

# Alternative fallback if sg3-utils unavailable:
for host in /sys/class/scsi_host/host*/scan; do echo '- - -' > $host 2>/dev/null; done
for port in /sys/class/fc_host/host*/issue_lip; do echo 1 > $port 2>/dev/null; done

6.1 Verify Multipath Devices Appeared

# List multipath devices
multipath -ll

# Check for your WWN (use your actual WWN prefix)
ls /dev/mapper/ | grep 3624a9370

# Should see device WITHOUT partition suffixes (due to kpartx blocking):
# Example: 3624a93708eabcb40cc4241b208501082
# NOT:     3624a93708eabcb40cc4241b208501082-part1

6.2 Verify No LVM VGs Activated

# Should NOT show VGs from source VM
vgs

# Should NOT show source VM's LVs
lvs

7.1 Get Next Available VMID

pvesh get /cluster/nextid
# Returns next available VMID, e.g., 108

7.2 Create VM with Raw Disk

# Create VM with basic settings
# Replace these example values with your actual values
VMID=108                                      # From Step 7.1
VM_NAME="myvm"                                # Your VM name
NODE="proxmox-02"                             # Target Proxmox node
NETWORK="vmbr0"                               # Your network bridge
WWN="3624a93708eabcb40cc4241b208501082"       # Your actual WWN from Step 3
SCSIHW="pvscsi"                               # See Step 7.2.1 for mapping

# Create VM
# --ostype values: l26 = Linux 2.6+, win10 = Windows 10/Server 2016+, win11 = Windows 11/Server 2022
qm create $VMID \
    --name "$VM_NAME" \
    --memory 8192 \
    --cores 4 \
    --sockets 1 \
    --cpu host \
    --scsihw $SCSIHW \
    --scsi0 "/dev/mapper/${WWN}" \
    --net0 "vmxnet3,bridge=${NETWORK}" \
    --ostype l26 \
    --boot order=scsi0

ostype reference: l26 = Linux (kernel 2.6 or newer), win10 = Windows 10/Server 2016+, win11 = Windows 11/Server 2022+, other = generic. This optimizes QEMU settings for the guest OS. Match this to the source VM's OS.

7.2.1 SCSI Controller Type Mapping

The --scsihw parameter specifies the SCSI controller type. Match this to your source VMware VM's SCSI controller to ensure the guest OS has compatible drivers already installed.

VMware to Proxmox SCSI Controller Mapping:

How to find your source VM's SCSI controller in VMware:

# Using govc (VMware CLI)
govc device.info -vm "VM_NAME" | grep -A5 "SCSI controller"

# Or in vCenter UI:
# VM -> Edit Settings -> Virtual Hardware -> SCSI Controller -> Type

What happens if you use the wrong controller?

• Wrong controller = guest won't boot (no driver for the disk controller)

• The guest OS must have drivers for the SCSI controller type you choose

• VMware VMs with PVSCSI have the pvscsi driver installed, so use pvscsi in Proxmox

• If unsure and the source was a modern VMware VM, pvscsi is the safest choice

Available Proxmox SCSI controller types:

7.3 For EFI VMs (from VMware EFI guests)

# Add EFI disk and BIOS type
qm set $VMID --bios ovmf
# Change "local-lvm" to your destination storage.
qm set $VMID --efidisk0 local-lvm:1,efitype=4m,pre-enrolled-keys=0

7.4 Preserve MAC Address (Optional but Recommended)

# Set MAC address from source VM. Replace the example MAC with your actual MAC from the source VM.
# Replace the example network adapter type with the actual type from the source VM or the desired type for the new VM. The example uses vmxnet3. Other common types are e1000 and virtio.
qm set $VMID --net0 "vmxnet3,bridge=${NETWORK},macaddr=00:50:56:93:a0:00"

7.5 Multiple Disks

For VMs with multiple disks, add each disk:

qm set $VMID --scsi1 "/dev/mapper/${WWN2}"
qm set $VMID --scsi2 "/dev/mapper/${WWN3}"

qm start $VMID

# Verify running
qm status $VMID

At this point, the VM is running from the raw multipath devices. The next step migrates the disks to Proxmox-managed storage.

This step performs a live block copy from the raw /dev/mapper device to LVM-backed Proxmox storage using QEMU's QMP interface. The Proxmox GUI/API is not able to perform this operation due to the direct mapping of the storage in the VM config. An error will occur if attempted stating the system is unable to parse the volume ID as Proxmox expects the volume ID to be in the format of storage:vm-<vmid>-disk-<N>.

9.1 Identify Target Storage VG

# Check storage.cfg for your target storage
# Replace "iscsi" with your actual storage ID
cat /etc/pve/storage.cfg | grep -A5 "lvm: iscsi"

# Note the vgname, e.g., "proxmox-cluster-pool"

9.2 Get Source Disk Size

blockdev --getsize64 /dev/mapper/${WWN}
# Returns size in bytes, e.g., 107374182400 (100GB)

# Convert to MB for LV creation
SIZE_MB=$((107374182400 / 1024 / 1024))

9.3 Create Target LV

# This is the target VG for the destination VM disk
VGNAME="proxmox-cluster-pool"
LV_NAME="vm-${VMID}-disk-0"

lvcreate -L ${SIZE_MB}M -n ${LV_NAME} ${VGNAME} -y -Wy

# Verify LV exists
lvs ${VGNAME}/${LV_NAME}

9.4 Connect to QMP Socket

The VM must be running. Connect to its QMP socket:

# Verify socket exists
ls -la /run/qemu-server/${VMID}.qmp

9.5 Perform Live Mirror via QMP

QMP requires a persistent session - all commands must be sent within the same connection. Use socat in interactive mode to maintain the session.

9.5.1 Set Variables

# Replace these example values with your actual values
VMID=108                                      # Your VM ID from Step 7
WWN="3624a93708eabcb40cc4241b208501082"       # Your actual WWN from Step 3
VGNAME="proxmox-cluster-pool"        # Your target LVM volume group
LV_NAME="vm-${VMID}-disk-0"
LV_PATH="/dev/${VGNAME}/${LV_NAME}"
QMP_SOCK="/run/qemu-server/${VMID}.qmp"

9.5.2 Connect to QMP (Interactive Session)

Start an interactive session that stays open:

socat - UNIX-CONNECT:${QMP_SOCK}

You'll see a greeting message like:

{"QMP": {"version": {...}, "capabilities": ["oob"]}}

9.5.3 Negotiate Capabilities

Type this command (required before any other commands):

{"execute": "qmp_capabilities"}

Expected response: {"return": {}}

9.5.4 Query Named Block Nodes

Query the block device structure to identify the node names needed for the mirror operation.

{"execute": "query-named-block-nodes"}

Understanding the Output:

QEMU creates a multi-layer block device stack. You'll see multiple entries for the same physical device with 3 layers:

Layer 1 (Top):    drive-scsi0          [throttle driver]  <- Use for "device" parameter
                       |
Layer 2 (Middle): ffa5d4038f11ff...    [raw driver]       <- Use for "replaces" parameter
                       |
Layer 3 (Bottom): efa5d4038f11ff...    [host_device driver] -> /dev/mapper/3624a9370...

Identify the three node names:

1. Top throttle node - Look for:

   • "node-name": "drive-scsi0" (or drive-virtio0, drive-sata0, etc.)

   • "drv": "throttle"

   • "backing_file_depth": 1

   • Has a "children" array

2. Child below throttle (raw driver) - Look in the throttle node's children array:

   • Find the entry with "child": "file"

   • Note the "node-name" (will be a hash like ffa5d4038f11ff4c1cbb20961716981)

   • This node will have "drv": "raw"

3. Bottom host_device node - Look for:

   • "drv": "host_device"

   • "file": "/dev/mapper/3624a9370..." (your source device)

   • "children": [] (empty array)

   • Note the "node-name" (will be a hash like efa5d4038f11ff4c1cbb20961716981)

Example from output:

Top throttle node:

{
  "node-name": "drive-scsi0",
  "drv": "throttle",
  "backing_file_depth": 1,
  "children": [
    {
      "node-name": "ffa5d4038f11ff4c1cbb20961716981",
      "child": "file"
    }
  ],
  ...
}

Child below throttle (raw driver):

{
  "node-name": "ffa5d4038f11ff4c1cbb20961716981",
  "drv": "raw",
  "backing_file_depth": 0,
  "children": [
    {
      "node-name": "efa5d4038f11ff4c1cbb20961716981",
      "child": "file"
    }
  ],
  "file": "/dev/mapper/3624a93708eabcb40cc4241b208501082"
}

Bottom host_device node:

{
  "node-name": "efa5d4038f11ff4c1cbb20961716981",
  "drv": "host_device",
  "backing_file_depth": 0,
  "children": [],
  "file": "/dev/mapper/3624a93708eabcb40cc4241b208501082"
}

Write down these three node names - you'll need them in the following steps.

Common top-level device names:

9.5.5 Add Target Block Device

Add the target LVM volume as a block device node.

Replace the filename with your actual LVM path:

{"execute": "blockdev-add", "arguments": {"driver": "raw", "node-name": "mirror-target-scsi0", "file": {"driver": "host_device", "filename": "/dev/proxmox-cluster-pool/vm-108-disk-0"}}}

Parameters explained:

• "driver": "raw" - No format, just raw block device

• "node-name": "mirror-target-scsi0" - Temporary name for the target (must match disk number, e.g., mirror-target-scsi0, mirror-target-scsi1)

• "filename" - Path to your LVM logical volume (from step 9.3)

Expected response: {"return": {}}

9.5.6 Start the Block Mirror

Start the live block mirror operation using the node names from step 9.5.4.

You need TWO node names from step 9.5.4: 1. Top throttle node (e.g., drive-scsi0) - for the device parameter 2. Child below throttle (e.g., ffa5d4038f11ff...) - for the replaces parameter

{"execute": "blockdev-mirror", "arguments": {"job-id": "mirror-scsi0", "device": "drive-scsi0", "target": "mirror-target-scsi0", "sync": "full", "replaces": "ffa5d4038f11ff4c1cbb20961716981"}}

Parameters explained:

• "job-id": "mirror-scsi0" - Unique identifier for this mirror job (you choose this)

• "device": "drive-scsi0" - Top throttle node from step 9.5.4

• "target": "mirror-target-scsi0" - Node name you created in step 9.5.5

• "sync": "full" - Copy all data (full synchronization)

• "replaces": "ffa5d4038f11ff..." - Child below throttle (raw driver) from step 9.5.4

Critical: The replaces parameter must be the child node below the throttle (the one with "drv": "raw"), NOT the bottom host_device node.

Expected response: {"return": {}}

9.5.7 Monitor Mirror Progress

Run this command repeatedly to check progress:

{"execute": "query-block-jobs"}

Output shows progress:

{"return": [{"device": "mirror-scsi0", "len": 107374182400, "offset": 53687091200, "ready": false, ...}]}

• Calculate percentage: offset / len * 100

• Wait until "ready": true appears

9.5.8 Complete the Mirror (when ready=true)

When the mirror job shows "ready": true, complete the switchover:

{"execute": "block-job-complete", "arguments": {"device": "mirror-scsi0"}}

Parameters explained:

• "device": "mirror-scsi0" - The job-id from step 9.5.6

Expected response: {"return": {}}

Wait a few seconds for the job to finish.

9.5.9 Verify Job Completed

Verify the mirror job has completed:

{"execute": "query-block-jobs"}

Should return: {"return": []} (empty array means no jobs running).

The VM is now using the target device (LVM volume).

9.5.10 Verify Source Nodes Were Replaced (Optional)

You can optionally verify that the replaces parameter worked correctly by querying the block nodes again:

{"execute": "query-named-block-nodes"}

Look for the drive-scsi0 node. You should see:

• The children array now points to mirror-target-scsi0 instead of the old source node

• The file field shows the target device path (e.g., /dev/nvme-tcp/vm-105-disk-0)

• The old source nodes (e.g., ffa5d4038f11ff... and efa5d4038f11ff...) no longer exist

Example of successful replacement:

{
  "node-name": "drive-scsi0",
  "drv": "throttle",
  "children": [{"node-name": "mirror-target-scsi0", "child": "file"}],
  "file": "json:{...\"filename\": \"/dev/nvme-tcp/vm-105-disk-0\"...}"
}

What the replaces parameter did:

• Automatically swapped the source node with the target node in the block device graph

• Deleted the old source nodes (ffa5d4038f11ff... and efa5d4038f11ff...)

• Released file descriptors on the old source device

• The throttle node now points directly to the target device

9.5.11 Exit QMP Session

Exit the socat session:

Press Ctrl+C

9.5.12 Verify Device Released

In a separate terminal (outside the QMP session):

# Verify device is released - should show Open count: 0
dmsetup info ${WWN} | grep "Open count"

If the open count is 0, the source device has been successfully released and the VM is now running entirely on the target LVM volume.

9.6 Update VM Configuration

After the mirror completes, update the VM config to reference the LVM storage instead of the raw device:

# Edit VM config
nano /etc/pve/qemu-server/${VMID}.conf

# Change the scsi0 line from (example WWN shown):
#   scsi0: /dev/mapper/3624a93708eabcb40cc4241b208501082
# To (use your storage name and LV name):
#   scsi0: iscsi:vm-108-disk-0,size=100G

Where iscsi is your Proxmox storage name and vm-108-disk-0 is the LV name you created in Step 9.3.

After the live mirror completes and the VM config is updated, clean up the temporary resources.

10.1 Verify Device is No Longer in Use

# Check open count - should be 0
dmsetup info ${WWN} | grep "Open count"

# If still in use, identify what's holding it:
fuser -v /dev/mapper/${WWN}
lsof /dev/mapper/${WWN}

10.2 Remove Multipath Device from All Nodes

On ALL Proxmox nodes:

# Remove kpartx partition mappings (if any exist)
kpartx -d /dev/mapper/${WWN}

# Deactivate any LVM VGs on the device (if any were activated)
# First identify VGs:
pvs /dev/mapper/${WWN}*
# Then deactivate:
vgchange -an <vg-name>

# Remove multipath device
multipath -f ${WWN}

# Verify device is gone
ls /dev/mapper/ | grep ${WWN}  # Should return nothing

10.3 Detach Volume from Host Group on Everpure

Using purecli:

purevol disconnect <volume-name> --hgroup <host-group-name>

# Example:
purevol disconnect proxmox-myvm-disk-0 --hgroup proxmox-cluster

# Verify disconnected
purevol listconnection proxmox-myvm-disk-0

10.4 Delete Cloned Volume from Everpure

Using purecli:

# Destroy volume (moves to destroyed state)
purevol destroy <volume-name>

# Eradicate (permanently delete)
purevol eradicate <volume-name>

# Example:
purevol destroy proxmox-myvm-disk-0
purevol eradicate proxmox-myvm-disk-0

10.5 Remove LVM Filter and Udev Rules

On ALL Proxmox nodes:

# Remove WWN from filter list
grep -v "^${WWN}$" /var/lib/migrator/filtered_wwns > /tmp/filtered_wwns.tmp
mv /tmp/filtered_wwns.tmp /var/lib/migrator/filtered_wwns

# If filter list is now empty, remove the config files:
if [ ! -s /var/lib/migrator/filtered_wwns ]; then
    rm -f /etc/lvm/lvm.conf.d/pure-migrator.conf
    rm -f /etc/udev/rules.d/59-migrator-skip-kpartx.rules
    udevadm control --reload-rules
fi

# If other WWNs remain, rebuild the filter files manually
# (See Step 2 for format)