> For AI agents: the complete documentation index is available at /en/llms.txt, the full documentation bundle is available at /en/llms-full.txt.

# Sync from your own PBS

If you already have your own Proxmox Backup Server on site, you can use
GetGeek as an **offsite copy** instead of pointing your Proxmox VE directly
at us. Your local PBS then sends its backups to your datastore with us
using a **push sync job**, on a schedule you choose. That is the third copy
in another location that the 3-2-1 rule requires, and your existing backup
jobs in PVE do not need to change.

**Requirement:** Proxmox Backup Server **3.3 or later** (the push direction
arrived in 3.3). Some options below require newer versions and are marked
accordingly.

Allow 15 minutes for the setup. The first sync sends all data and can take
hours; after that it is incremental.

## Step 1: Prepare your datastore with us \{#steg-1-forbered-ditt-datastore-hos-oss}

Log in at **[https://sto-pbs.getgeek.se:8007](https://sto-pbs.getgeek.se:8007)** with your **password**.

### A dedicated namespace for the sync \{#en-egen-namespace-for-synken}

Create a namespace that only the push job writes to, for example
`kluster1` or the name of your local PBS:
**Datastore → `<datastore>` → Content → Add Namespace**
(see [Namespaces](/en/proxmox-backup/kom-igang.md#namespaces)).

This keeps the copies from your PBS apart from any backups your PVE makes
directly to us. The sync job only writes to groups owned by its own token.
A group already owned by someone else, for example your PVE's token, is
skipped with the warning _not owned by remote user … skipping upload_, and
those backups never arrive.

The token cannot create namespaces. If you sync a local namespace structure
with several levels, the corresponding namespaces must exist on our side
before the job runs.

### A dedicated API token for the sync \{#en-egen-api-token-for-synken}

Create a **new token just for this job**, exactly as in
[Getting started, step 1](/en/proxmox-backup/kom-igang.md#steg-1-skapa-en-api-token-at-din-proxmox):

- **Token Name:** e.g. `push-lokal`
- **Path:** `/datastore/<datastore>` (or `/datastore/<datastore>/<namespace>`)
- **Role:** `DatastoreBackup`, **Propagate** checked

:::warning One token per push job, and nothing other than `DatastoreBackup`
Everything a push job sends is owned by the token in the remote
configuration. If two jobs share the same token, they can delete or skip
each other's backups. Proxmox therefore recommends a separate user or
token per push job.

With `DatastoreBackup` your local PBS can add backups with us, but not
delete them. If the local server is compromised, the offsite copy is still
untouched, and that is the whole point of it.
:::

## Step 2: Encryption — decide before the first sync \{#steg-2-kryptering-bestam-dig-innan-forsta-synken}

A push job sends your backups **as they are stored on your PBS**. If they
are unencrypted locally, they end up unencrypted with us. The transfer is
always encrypted (TLS), but the storage is not. Choose one of three
approaches:

| Mode                                        | How                                                                                                                                                                              | Requires |
| ------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------- |
| **A. Encrypt already in PVE** (recommended) | The storage in your PVE has an encryption key, see [Encryption & key](/en/proxmox-backup/kryptering.md). The backups are already encrypted on your PBS and are sent as they are. | —        |
| **B. Send only encrypted**                  | Check **Encrypted Only** in the sync job. Unencrypted backups are skipped instead of sent. A safeguard in case someone adds a VM without a key.                                  | PBS 3.4+ |
| **C. Let the sync job encrypt**             | Your PBS encrypts unencrypted backups with its own key before they are sent. Already encrypted ones are sent unchanged.                                                          | PBS 4.2+ |

Combining **A** and **B** is a good idea. Mode **C** suits cases where your
local backups for some reason have to be unencrypted.

:::danger The key stays with you — lose it and the backup is gone
We never have your encryption keys and cannot recover encrypted data
without them. That applies both to the PVE key (mode A) and the sync job's
key (mode C). Back up the key **before** the first sync, in a place that is
not your PBS.
:::

**Mode C — create the key on your PBS:** **Configuration → Encryption Keys →
Add → Add Sync Encryption Key**. Give it a name, let it be generated
automatically and export it straight away.

![The Edit: Encryption Key dialog](/_assets/image/sync-key-add.5dcf87887e.png)
## Step 3: Add GetGeek as a remote on your PBS \{#steg-3-lagg-till-getgeek-som-remote-pa-din-pbs}

On **your** PBS: **Configuration → Remotes → Add**

- **Remote ID:** any name, e.g. `getgeek`
- **Host:** `sto-pbs.getgeek.se` (the port is the default, `8007`)
- **Auth ID:** `<username>@pbs!push-lokal` (the token from step 1)
- **Password:** the token's secret
- **Fingerprint:** leave empty

![The Add: Remote dialog](/_assets/image/remote-add.7218ea173f.png)
Our certificate comes from Let's Encrypt and is accepted automatically, so
no fingerprint is needed.

## Step 4: Create the push job \{#steg-4-skapa-push-jobbet}

On **your** PBS: **Datastore → `<your local datastore>` → Sync Jobs →
Add → Add Push Sync Job**.

**The Options tab:**

- **Local Namespace:** what you want to send (`Root` = everything)
- **Sync Schedule:** after your local backups have finished, e.g. `02:00`
- **Target Remote:** `getgeek`
- **Target Datastore:** `<datastore>`
- **Target Namespace:** the namespace from step 1
- **Remove Vanished:** **unchecked** — the token is not allowed to delete,
  and that is intentional

![The Options tab in Add: Sync Job - Push Direction](/_assets/image/sync-push-options.7cdb6e6f81.png)
**Check Advanced** for more options:

- **Encrypted Only:** mode B from step 2
- **Transfer Last:** how many of the most recent backups per VM are sent.
  See the tip below.
- **Verified Only:** send only backups that have passed verify on your PBS

![The Advanced fields with Encrypted Only checked](/_assets/image/sync-push-advanced.bc71646140.png)
**The Encryption tab** (mode C only): choose the key from step 2 as the
**Active Encryption Key**.

![The Encryption tab with an active key](/_assets/image/sync-push-encryption.dbf3011cc9.png)
If you only want to send certain VMs: use the **Group Filter** tab.

:::tip Do not send history that will be pruned anyway
Our [retention](/en/proxmox-backup/retention.md) applies to your entire datastore, including
the namespace the push job writes to. If your local PBS holds a year of
backups, the first sync sends all of them, and the prune job that same
evening removes whatever falls outside the policy. Set **Transfer Last** to
something reasonable (e.g. `14`) to save time and bandwidth on the first
sync. If you want a longer history offsite than the default: get in touch
and we will change the retention for your datastore.
:::

## Step 5: Run and check \{#steg-5-kor-och-kontrollera}

1. Select the job and click **Run now**. Follow the progress under
   **Show Log**.
2. Log in on our side and open **Datastore → `<datastore>` → Content**.
   The groups should be under the namespace you chose.
3. Test a restore from the copy with us, exactly as in
   [Getting started, step 4](/en/proxmox-backup/kom-igang.md#steg-4-testa-aterlasning-gor-det-nu-inte-nar-det-brinner).
   You can add our datastore as additional storage in a PVE with the usual
   token and restore from there.

After the first sync, the job only sends new backups. Each backup appears
on our side as a separate _backup_ task with its original timestamp, so
many consecutive tasks during the first sync is normal.

## Troubleshooting \{#felsokning}

| Error                                                               | Cause                                                                                                                                |
| ------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------ |
| `permission check failed` right after the token was created         | Permissions take effect after a few seconds. Wait and run it again.                                                                  |
| `permission check failed` on a namespace                            | The namespace does not exist on our side, or the token's **Path** does not cover it. Create it with your password.                   |
| _Group '…' not owned by remote user '…' on target, skipping upload_ | The group on our side is owned by another token, e.g. your PVE's. Use a dedicated namespace for the push job.                        |
| Backups are skipped                                                 | With **Encrypted Only**, unencrypted backups are skipped. Partially encrypted backups are also skipped in mode C. See the job's log. |

## Support \{#support}

Questions or problems? Email **[hej@getgeek.se](mailto:hej@getgeek.se)** — the same engineers who
run the platform will answer.
