Bash script to transfer or clone LXC containers and VMs between Proxmox VE nodes — or on the same node — without requiring a cluster. Workflow: stop → vzdump backup → transfer → restore → start. The guest type (lxc / qemu) is detected automatically.
- Works for LXC containers and QEMU VMs — type is auto-detected
- Four modes depending on which hosts you specify (see below): remote→remote, push, pull, local clone
- No Proxmox cluster required — plain SSH is enough
- Streaming transfer (no intermediate file on the machine running the script), optional progress bar via
pv - Size verification after transfer
- Handles unprivileged containers correctly (dump directory permissions are fixed automatically)
- Safety checks: SSH connectivity, source ID exists, target ID free, target storage exists
- Automatic cleanup of temporary files on success and on failure
- Key=value arguments in any order, interactive prompts for missing credentials
The side you omit runs on the local node (the machine executing the script):
| Arguments given | Mode | Behavior |
|---|---|---|
s= and d= |
remote → remote | Transfer between two remote nodes |
only d= |
push | Back up local guest, restore it on d= |
only s= |
pull | Back up guest on s=, restore it locally |
| none | local clone | Copy a guest to a new ID on the same node |
- Proxmox VE on all involved nodes (
vzdump,pct/qm,pvesm) - SSH access with root privileges on remote nodes
sshpasson the executing machine (only for password authentication)pv(optional, shows a progress bar during transfer)- Enough free space in
/var/tmpon source and destination for the compressed backup
wget https://raw.githubusercontent.com/sorglos-it/pve-transfer/main/pve-transfer.sh
chmod +x pve-transfer.sh./pve-transfer.sh [s=<src-host>] [d=<dst-host>] oid=<source-id> nid=<target-id> st=<storage> [u=<user>] [p=<pass>]
| Key | Alias | Description |
|---|---|---|
s= |
src= |
Source PVE host (omitted = local node) |
d= |
dst= |
Destination PVE host (omitted = local node) |
oid= |
– | Source VMID / CTID |
nid= |
– | New VMID / CTID on the destination |
st= |
storage= |
Target storage on the destination |
u= |
user= |
SSH user (optional, prompted if a remote side is used) |
p= |
pass= |
SSH password (optional, prompted if missing; empty = key auth) |
Argument order is arbitrary. -h, -help, --help or no arguments print the built-in help.
# Remote -> remote: move guest 100 from pve-1 to pve-2 as 102
./pve-transfer.sh s=pve-1.example.com d=pve-2.example.com oid=100 nid=102 st=ssd_1tb u=root p=secret
# Push: copy local guest 100 to pve-2 as 102
./pve-transfer.sh d=pve-2.example.com oid=100 nid=102 st=ssd_1tb
# Pull: fetch guest 100 from pve-1 and restore it locally as 102
./pve-transfer.sh s=pve-1.example.com oid=100 nid=102 st=local-lvm
# Local clone: copy guest 100 to ID 110 on this node
./pve-transfer.sh oid=100 nid=110 st=local-lvm- Checks — SSH connectivity, guest type detection, target ID free, target storage exists
- Stop — graceful shutdown (120 s timeout), hard stop as fallback
- Backup —
vzdump --mode stop --compress zstdinto a temporary directory - Transfer — archive is streamed source → destination through the executing machine (skipped for local clones)
- Restore —
pct restore/qmrestorewith the new ID on the chosen storage - Start — the restored guest is started on the destination
- The source guest stays stopped after a successful run. The backup preserves MAC addresses and IP configuration — running both copies simultaneously causes conflicts. Adjust the network config of the copy (
pct set/qm set) before starting both. - Passing
p=on the command line exposes the password in the process list and shell history. Prefer the interactive prompt or SSH key authentication. - In remote→remote mode the data is streamed through the machine running the script — consider its bandwidth, or run the script directly on one of the nodes.
- Exit codes:
0= success,1= error,2= usage.
If this script saved you time, you can support further development:
MIT