Migrate virtual machines between different Proxmox VM clusters
Migrate VMs between different Proxmox VE clusters.
Migrating a virtual machine (VM) on a PVE-cluster from one node to another is implemented in the Proxmox Virtual Environment (PVE). But migrating a VM from one PVE-cluster to another is not.
proxmove helps you move VMs between PVE-clusters with minimal hassle.
usage: proxmove [-h] [-c FILENAME] [-n] [--bwlimit MBPS] [--skip-disks] [--skip-start] [--version] source destination nodeid storage vm [vm ...] Migrate VMs from one Proxmox cluster to another. positional arguments: source alias of source cluster destination alias of destination cluster nodeid node on destination cluster storage storage on destination node vm one or more VMs (guests) to move optional arguments: -h, --help show this help message and exit -c FILENAME, --config FILENAME use alternate configuration inifile -n, --dry-run stop before doing any writes --bwlimit MBPS limit bandwidth in Mbit/s --skip-disks do the move, but skip copying of the disks; implies --skip-start --skip-start do the move, but do not start the new instance --version show program's version number and exit Cluster aliases and storage locations should be defined in ~/.proxmoverc (or see -c option). See the example proxmoverc.sample. It requires [pve:CLUSTER_ALIAS] sections for the proxmox "api" URL and [storage:CLUSTER_ALIAS:STORAGE_NAME] sections with "ssh", "path" and "temp" settings.
First you need to configure ~/.proxmoverc; see below.
When configured, you can do something like this:
$ proxmove banana-cluster the-new-cluster node2 node2-ssd the-vm-to-move 12:12:27: Attempt moving banana-cluster<e1400248> => the-new-cluster<6669ad2c> (node 'node2'): the-vm-to-move 12:12:27: - source VM the-vm-to-move@node1<qemu/565/running> 12:12:27: - storage 'ide2': None,media=cdrom (host=<unknown>, guest=<unknown>) 12:12:27: - storage 'virtio0': sharedsan:565/vm-565-disk-1.qcow2,format=qcow2,iops_rd=4000,iops_wr=500,size=50G (host=37.7GiB, guest=50.0GiB) 12:12:27: Creating new VM 'the-vm-to-move' on 'the-new-cluster', node 'node2' 12:12:27: - created new VM 'the-vm-to-move--CREATING' as UPID:node2:00005977:1F4D78F4:57C55C0B:qmcreate:126:user@pve:; waiting for it to show up 12:12:34: - created new VM 'the-vm-to-move--CREATING': the-vm-to-move--CREATING@node2<qemu/126/stopped> 12:12:34: Stopping VM the-vm-to-move@node1<qemu/565/running> 12:12:42: - stopped VM the-vm-to-move@node1<qemu/565/stopped> 12:12:42: Ejected (cdrom?) volume 'ide2' (none) added to the-vm-to-move--CREATING@node2<qemu/126/stopped> 12:12:42: Begin copy of 'virtio0' (sharedsan:565/vm-565-disk-1.qcow2,format=qcow2,iops_rd=4000,iops_wr=500,size=50G) to local-ssd 12:12:42: scp(1) copy from '/pool0/san/images/565/vm-565-disk-1.qcow2' (on sharedsan) to 'email@example.com:/node2-ssd/temp/temp-proxmove/vm-126-virtio0' Warning: Permanently added 'node2.the-new-cluster.com' (ECDSA) to the list of known hosts. vm-565-disk-1.qcow2 100% 50GB 90.5MB/s 09:26 Connection to san.banana-cluster.com closed. 12:22:08: Temp data '/node2-ssd/temp/temp-proxmove/vm-126-virtio0' on local-ssd 12:22:08: Writing data from temp '/node2-ssd/temp/temp-proxmove/vm-126-virtio0' to '/dev/zvol/node2-ssd/vm-126-virtio0' (on local-ssd) (100.00/100%) Connection to node2.the-new-cluster.com closed. 12:24:25: Removing temp '/node2-ssd/temp/temp-proxmove/vm-126-virtio0' (on local-ssd) 12:24:26: Starting VM the-vm-to-move@node2<qemu/126/stopped> 12:24:27: - started VM the-vm-to-move@node2<qemu/126/running> 12:24:27: Completed moving banana-cluster<e1400248> => the-new-cluster<6669ad2c> (node 'node2'): the-vm-to-move
Before, the-vm-to-move was running on banana-cluster on node1.
Afterwards, the-vm-to-move is running on the-new-cluster on node2. The the-vm-to-move on the banana-cluster has been stopped and renamed to the-vm-to-move--MIGRATED.
Set up the ~/.proxmoverc config file. First you need to define which clusters you have. For example banana-cluster and the-new-cluster.
; Example cluster named "banana-cluster" with 3 storage devices, one ; shared, and two which exist on a single node only. [pve:banana-cluster] api=https://user@pve:PASSWORD@banana-cluster.com:443 ; Example cluster named "the-new-cluster" with 2 storage devices; both ; storage devices exist on the respective nodes only. [pve:the-new-cluster] api=https://user@pve:PASSWORD@the-new-cluster.com:443
Next, it needs configuration for the storage devices. They are expected to be reachable over SSH; both from the caller and from each other (using SSH-agent forwarding).
The following defines two storage devices for the banana-cluster, one shared and one local to node1 only.
If on sharedsan, the images are probably called something like /pool0/san/images/VMID/vm-VMID-disk1.qcow2, while in Proxmox, they are referred to as sharedsan:VMID/vm-VMID-disk1.qcow2.
[storage:banana-cluster:sharedsan] ; "sharedsan" is available on all nodes firstname.lastname@example.org path=/pool0/san/images temp=/pool0/san/private [storage:banana-cluster:local@node1] ; local disk on node1 only email@example.com path=/srv/images temp=/srv/temp
If you use ZFS storage on the-new-cluster, the storage bits could look like this. Disk volumes exist on the ZFS filesystem node1-ssd/images and node2-ssd/images on the nodes node1 and node2 respectively.
Note that the temp= path is always a regular path.
[storage:the-new-cluster:node1-ssd@node1] firstname.lastname@example.org path=zfs:node1-ssd/images temp=/node1-ssd/temp [storage:the-new-cluster:node2-ssd@node2] email@example.com path=zfs:node2-ssd/images temp=/node2-ssd/temp
The config file looks better with indentation. The author suggests this layout:
[pve:banana-cluster] ... [storage:banana-cluster:sharedsan] ... [storage:banana-cluster:local@node1] ... [pve:the-new-cluster] ... [storage:the-new-cluster:node1-ssd@node1] ...
proxmove is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, version 3 or any later version.
0.1.0 - 2018-11-22
- Show that not seeing a VM is probably a PVEAdmin-permissions issue.
- Don’t die if image_size cannot be determined.
- Place the sample files and docs in a /usr/share/…proxmove subdir.
0.0.9 - 2017-03-28
- Added –no-verify-ssl option.
- Fix str-cast bug with ZFS destination creation.
- Fix ignoring of non-volume properties like “scsihw”.
0.0.8 - 2017-01-26
- Partial LXC container move implemented. Not complete.
- Allow ZFS paths to be more than just the a pool name. Instead of e.g. path=zfs:mc9-8-ssd we now also allow path=zfs:rpool/data/images. Closes #7.
v0.0.7 - 2016-10-07
- Instead of trusting on the “size=XXG” which may or may not be present in the storage volume config, it reads the QCOW header or ZFS volume size directly. Also checks that the values are available before attempting a move.
v0.0.6 - 2016-09-21
- Add –bwlimit in Mbit/s to limit bandwidth during transfer. Will use the scp(1) -l option or for ZFS use the mbuffer(1) auxiliary. As an added bonus mbuffer may improve ZFS send/recv speed through buffering. Closes #4.
- Add –skip-disks option to skip copying of the disks. Use this if you want to copy the disks manually. Closes #3.
- Add –skip-start option to skip autostarting of the VM.
- Adds optional pv(1) pipe viewer progress bar to get ETA in ZFS transfers.
- Add hidden –debug option for more verbosity.
- Add hidden –ignore-exists option that allows you to test moves between the same cluster by creating an alias (second config).
- Format is not always specified in the properties. If it isn’t, use the image filename suffix when available.
- Sometimes old values aren’t available in the “pending” list. Don’t croak. Closes #2.
- Begun refactoring. Testing bettercodehub.com.
- Also check whether temporary (renamed) VMs exist before starting.
v0.0.5 - 2016-08-30
- Added support for ZFS to ZFS disk copy. QCOW2 to ZFS and ZFS to ZFS is now tested.
v0.0.4 - 2016-08-30
- Major overhaul of configuration format and other changes.
- On missing disk (bad config), the zfs send stalls and mbuffer waits for infinity.
- Rename “VM” to “Instance” so “LXC” containers don’t feel misrepresented.
- Communicate with the storage API to check/configure storage (more easily).
- Create a --config command to set up a basic configuration. Combine with storage API reading.
- Fix so LXC-copying works (this is a bit tricky because of the necessity for a temporary image/tarball to install).
- Create a proxmoxer_api example that returns test data, so we can run tests.
- Let it work with pve-zsync – a daemon that syncs data between nodes: https://pve.proxmox.com/wiki/PVE-zsync