TL;DR:
The team at FactorFX built a Proxmox integration for Plakar that wraps Proxmox’s native vzdump backups and stores them as deduplicated Plakar snapshots, making VM and container backups portable, encrypted, and easy to restore across clusters.
Proxmox support had been requested many times… but the interesting part is that we didn’t end up building it ourselves.
Throughout 2025, the topic of virtual machine backups came up every now and then.
We knew we wanted to support it, but we weren’t convinced it was the highest priority. There are many things people want to back up, and virtual machines didn’t seem more urgent than S3, GCS or PostgreSQL in daily discussions.
Then something interesting happened.
Mathieu and I were holding a booth at the Capitole du Libre 2025, in Toulouse, and we kept getting the same question from visitors:
Can Plakar back up Proxmox clusters?
So many people asked that Mathieu started looking into how Proxmox backups worked.
Later, François-Xavier, who was also holding a booth for FactorFX, came over and told me:
Hey Gilles, we need to talk about Proxmox support in Plakar!
A couple of weeks later, at the Tech Rocks event, the exact same scenario unfolded with a completely different crowd.
At that point, it became clear this was no longer a “should we prioritize Proxmox” debate, but a “who does it and when”.
FactorFX enters the game !
Technically, our team can write most integrations fairly quickly, anywhere between an hour and a few days depending on the complexity.
Which makes it tempting to just do them ourselves: spend a few hours on it and the problem is solved.
But that’s not what we want.
Our goal is for Plakar to become a de-facto standard to back up anything, and that can’t happen if we’re the only ones writing integrations. The ecosystem of tools out there is simply too large.
Instead, we want a layered ecosystem:
- official integrations maintained by us
- third-party integrations that we review and stamp as “trusted by Plakar”
- community integrations where anyone can build support for whatever software they like
Since FactorFX had already shown interest in Proxmox, this felt like the perfect opportunity to bootstrap that model.
After a few discussions, Gilles Dubois came back a few days later with a working Proxmox integration !
What is Proxmox ?
Proxmox Virtual Environment (PVE) is an open-source virtualization platform used to run and manage virtual machines and containers.
At a technical level, it combines several well-known pieces of infrastructure software:
- KVM to run virtual machines
- LXC to run containers
- multiple storage backends such as ZFS, Ceph or simple directory storage
- a web UI and REST API to manage everything
Proxmox includes many of the features people expect from enterprise virtualization platforms, including live migration, high availability, storage replication, built-in backups and clustering. All of this in a single system that is relatively easy to deploy and operate.
Over the past year, it has also become a very popular alternative to other hypervisors, especially as many organizations started reconsidering their virtualization stack following the VMware/Broadcom flustercluck.
Which explains why, every time we showed Plakar at a conference booth, someone eventually asks:
That’s cool… but can it back up Proxmox?
Why use Plakar for Proxmox backups?
Proxmox already includes a backup tool called vzdump,
and it works very well,
so why introduce another tool in the mix?
The answer is that Plakar does not replace Proxmox backups, it extends them.
The integration simply relies on vzdump to generate the backup archives,
and then stores them inside Plakar snapshots.
This means the backups behave exactly like native Proxmox backups,
while gaining a few extra properties along the way.
For example, Plakar deduplicates data across snapshots. If multiple virtual machines share the same base image, that data only needs to be stored once.
Snapshots can also be archived to different storage backends, making it easy to keep long-term backups on object storage or cold storage.
Finally, because Plakar integrations share the same connector model, data is not tied to a single environment. A VM backed up from Proxmox could later be restored to another cluster, archived elsewhere, or inspected without restoring the entire machine.
Installing the Proxmox integration
The proxmox integration has been committed to a public repository and is only available for plakar starting with v1.1.0-beta.
To test it, you first need to install our latest beta of plakar:
$ go install github.com/PlakarKorp/plakar@v1.1.0-beta.7
You can then either use our prebuilt package by authenticating to our platform:
$ plakar login
[...]
$ plakar pkg add proxmox
$
Or build the integration yourself…
$ plakar pkg build proxmox
/usr/bin/make -C /var/folders/9x/9k0f6mc10sbd0_kfx63__fvc0000gn/T/build-proxmox-v1.1.0-rc.1-4157532844
83b7da91: OK ✓ /
83b7da91: OK ✓ /manifest.yaml
83b7da91: OK ✓ /proxmoxExporter
83b7da91: OK ✓ /proxmoxImporter
Plugin created successfully: proxmox_v1.1.0-rc.1_darwin_arm64.ptar
… and install the resulting ptar:
$ plakar pkg add ./proxmox_v1.1.0-rc.1_darwin_arm64.ptar
Aaaaaand that’s it.
Local vs remote operation
Before showing how it’s used, a few words about how it works.
The integration supports two operating modes.
In local mode, Plakar runs directly on the Proxmox node:
Proxmox node
├ vzdump
└ plakar
This is the simplest setup.
In remote mode, Plakar runs on a separate machine and connects to the Proxmox host via SSH:
Backup server
│
│ SSH
▼
Proxmox node
└ vzdump
This allows a single Plakar instance to back up multiple hypervisors.
Backing up virtual machines and containers
Once the integration is installed, backing up Proxmox virtual machines and containers becomes straightforward.
First, we configure a Proxmox source:
$ plakar source add myProxmox proxmox+backup://10.0.0.10 \
mode=remote \
conn_username=root \
conn_identity_file=/path/to/key \
conn_method=identity
Then we can start backing up workloads.
For example, to back up a single virtual machine:
$ plakar backup -o vmid=101 @myProxmox
Or all the machines in a pool:
$ plakar backup -o pool=prod @myProxmox
Or even the entire hypervisor:
$ plakar backup -o all @myProxmox
Under the hood,
the integration invokes vzdump,
collects the resulting archive,
and ingests it into a Plakar snapshot.
Once stored, the backup benefits from all Plakar features such as deduplication, encryption and snapshot browsing.
Restoring virtual machines and containers
Restoring workloads is equally straightforward.
First, configure a Proxmox destination:
$ plakar destination add myProxmox \
proxmox+backup://10.0.0.10 \
mode=remote \
conn_username=root \
conn_identity_file=/path/to/key \
conn_method=identity
Then restore a snapshot:
$ plakar restore -to @myProxmox <snapid>
The integration uploads the dump archive to the Proxmox node and restores it using native tools:
qmrestorefor virtual machinespct restorefor containers
It is also possible to restore only one VM from a snapshot containing multiple machines:
$ plakar restore -to @myProxmox <snapid>:/backup/qemu/101_myvm
If configured, the restored machine can automatically start once the restore completes.
Beyond simple backups
Because Plakar integrations share the same connector model, data is not locked into a single environment.
For example, a virtual machine backed up from Proxmox could be:
- inspected with high granularity using the Plakar UI.
- stored in
a minio instance, err, an S3 bucket at Scaleway, OVH or Exoscale. - synchronized between stores at Scaleway, OVH and Exoscale for multiple copies.
- exported to a ptar archive and archived in cold storage.
- restored to another cluster.
This flexibility enables backup workflows that go far beyond traditional hypervisor backups.
Wrapping up
This Proxmox integration is still early, but it’s already working well.
If you’re running a Proxmox cluster, be it on-premise or in the cloud, please give it a try and let us know what you think!
And if you’re interested in writing integrations yourself, take a look at what the FactorFX team achieved: they had an initial version working in just a few days, with no prior experience whatsoever with our codebase.
We can help you get started and bootstrap things, and you might just end up building that one integration that everyone wants but nobody has written yet !