Set a backup schedule and retention on Dynamic.

Where this happens on Dynamic

Everything in this guide is configured in Proxmox VE — the schedule and the keep-* retention flags live on your backup job. You don't need (and don't have) the PBS web interface on this plan: pruning to your keep rules and the garbage collection that actually frees space run on our side, honoring exactly what you set in the job. You declare the policy; we operate it.

Step 1: Create the backup job

In the Proxmox VE interface: Datacenter → Backup → Add. The choices that matter:

• Selection — start with All guests rather than a hand-picked list. The most expensive backup failure is the VM that quietly never joined the job. Exclude the few you truly don't want.
• Storage — your Dynamic PBS storage (added in the connect guide).
• Mode — Snapshot. Guests keep running while they're backed up.
• Compression — ZSTD. Fast, and smaller stored data is a smaller Dynamic bill.

Step 2: Pick a schedule

The Schedule field takes a calendar event. Three patterns cover most environments:

# nightly at 1 AM — the right default for most
01:00
# weeknights only, at 10:30 PM
mon..fri 22:30
# every 6 hours, for guests that change fast
00/6:00

After the first full upload, backups are incremental — usually minutes, not hours — so a nightly schedule costs far less bandwidth than it sounds. If you run multiple jobs, stagger their start times; planning backup windows covers how long jobs really take.

Step 3: Set retention — the part that is also your bill

On the job's Retention tab, the keep-* fields decide how much history survives. A default that works for most teams:

keep-last    3    # the 3 most recent, no matter what
keep-daily   7    # one per day, a week back
keep-weekly  4    # one per week, a month back
keep-monthly 6    # one per month, half a year back

That's roughly 20 snapshots per guest — but because PBS deduplicates between snapshots, 20 snapshots costs nowhere near 20× the data. Only changed chunks are stored, and Dynamic bills on that compressed, deduplicated size. Longer retention costs less than you'd guess; unlimited retention still isn't a policy.

Leaving every keep field empty means nothing is ever removed — your history and your bill grow forever. Set the flags. For designing a policy against actual compliance requirements, see designing retention schedules.

Step 4: Run it once and confirm

Select the job and click Run now. Confirm the task ends in TASK OK, then check the storage's Backups view — every guest in the job should have a fresh snapshot. While you're in the job dialog, set a notification email so failures reach a human instead of a log file.

One thing not to be surprised by: when retention removes old snapshots, your usage graph doesn't drop the same hour. Garbage collection reclaims the freed chunks on our schedule (PBS holds recently-touched chunks for about a day as a safety window) — the space and the bill follow your policy with a short lag.