How it works
Four steps to a backup you have actually restored from.
Proxmox Backup Server is free software. What you are buying is the operations around it — sizing, scheduling, verification, garbage collection, offsite copies, and the restore drill nobody otherwise gets round to.
Connecting your cluster
Onboarding is deliberately boring. Nothing here requires a maintenance window on your hypervisor.
Add the datastore as storage in PVE
Step 1In Datacenter → Storage you add a Proxmox Backup Server target. We give you the server address, the datastore name, and the TLS fingerprint. The fingerprint matters: it is what stops your cluster trusting a server that merely claims to be ours.
Authenticate with an API token, not root
Step 2We issue a PBS API token scoped to your datastore. It can write backups and read them back for restores. Depending on the change-control model you choose, it does not carry prune or delete rights, so a compromised PVE host cannot erase its own backup history.
Schedule the jobs
Step 3Backup jobs are configured per VM or per pool with a retention policy attached. Where the WAN link is shared with production traffic, we rate-limit the job so the first full backup does not saturate the site's uplink during business hours.
First verify, then first restore
Step 4Once the first backup completes, a verify job re-reads the chunks rather than assuming the write succeeded. Then we restore one non-production VM and document it. Onboarding is not finished when a backup appears in the list; it is finished when something has been restored from it.
What we operate
This is the list that turns free software into a service. Each line is something that quietly breaks backups when nobody owns it.
- PBS version currency and OS patching on the backup server
- Datastore sizing, and expansion proposed before the datastore fills
- Disk and SMART health monitoring on the underlying storage
- Prune schedules and garbage-collection windows that do not collide with backups
- Verify job scheduling and failure alerting
- Sync jobs to the offsite or secondary datastore
- Backup of the PBS configuration itself, which is the thing people forget
- Certificate expiry monitoring on the PBS endpoint
What stays yours
- Your PVE cluster, unchanged — we do not take over the hypervisor
- Your encryption keys, if you enable client-side encryption
- The decision about what gets backed up and how long it is kept
- The ability to restore yourself at any time, without raising a ticket
Questions before onboarding
Why not just run PBS ourselves? It is free.
You absolutely can, and if you have the time it is a good product. What people actually pay for is the operations around it: sizing the datastore, keeping garbage collection out of the backup window, noticing that verify jobs started failing three weeks ago, handling encryption keys properly, holding a genuine offsite copy, and running restore drills that nobody ever has time for. The software is free. The discipline is the product.
Which ports need to be open?
PBS listens on TCP 8007. Your PVE hosts connect outbound to that port on the backup server; nothing needs to be opened inbound on your firewall towards your cluster. If you host the PBS on your own premises instead, the same port applies between your cluster and your backup server.
How long does the first backup take?
The first run is a full copy, so it is bounded by your uplink rather than by PBS. Subsequent runs are incremental with deduplication and are usually a small fraction of that. For large first seeds over a slow link we discuss seeding options during onboarding rather than letting a job run for a fortnight.
Will backups saturate our internet connection?
Not if the job is rate-limited, which is part of the standard configuration where the link is shared with production. We set the cap against your actual available bandwidth and the window you want jobs to run in.
Can we get our data out if we leave?
Yes. Your backups can be synced to a PBS instance you own, or exported. Ask for the exit terms in writing before you sign — from us or from any backup vendor. A service that is easy to join and vague about leaving is a service worth reading twice.
Start with the health check
Before any commercial conversation, we look at what your current backups are actually doing — job success, last verified restore, datastore fill, GC lag, and whether encryption is on. It costs nothing and it is useful even if you never buy anything.