> For clean Markdown of any page, append .md to the page URL. For a complete documentation index, see https://ibee.ai/docs/llms.txt (full text of every page: https://ibee.ai/docs/llms-full.txt). For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://ibee.ai/docs/_mcp/server. # VM backups > Automatic, scheduled copies of a Cloud VM or GPU VM. What's included, the schedule and how long copies are kept, how they're billed, what a restore overwrites, and how backups differ from snapshots. A **backup** is an automatic copy of a VM's disks that IBEE takes on a schedule you choose, daily or weekly. Each copy is called a **recovery point**. If the VM breaks, or someone deletes important files, you can put a recovery point back onto the VM, build a new VM from it, or restore a single disk. Backups are off until you turn them on for each VM, and they cost extra: you pay for the storage the recovery points use. ## When to use backups | You want… | Use | | ----------------------------------------------------- | -------------------------------------------------------------------------------- | | A copy every day or week without having to remember | **Backups** (this page) | | A copy right now, before an upgrade or a risky change | [Snapshots](/docs/tools/snapshots) | | A copy that's kept until you delete it | [Snapshots](/docs/tools/snapshots), or **Convert to Snapshot** on a backup | | Copies of files outside the VM, or in your own format | Your own tools, writing to [Object Storage](/docs/infrastructure/object-storage) | Most production VMs use both: scheduled backups as the safety net, and a snapshot before each risky change. ## How it works 1. You turn on **Automated Backups** on the VM's **Backups** tab and pick a schedule. 2. At the scheduled time (within 30 minutes after it), IBEE copies the root disk and **every volume attached at that moment**. The VM keeps running. 3. The first copy is a **full** backup. For the next 7 days, copies are **incremental**: they store only what changed. Then a new full backup starts the next 7-day chain. 4. Each copy is stored separately from the VM's own disks and appears as a recovery point. 5. Old recovery points are deleted automatically after **7 days**, a whole weekly chain at a time. ## Before you begin * A Cloud VM or GPU VM, and **write access** to its workspace. * Credit in your wallet: backup storage is billed per GB per month. * Data that must be **consistent** (databases) is best protected by stopping writes or using the database's own dump tool before the backup time. IBEE copies the disks without pausing the VM's filesystem, so a backup is like the disk after a sudden power cut. ## Key facts | Question | Answer | | ---------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | What's included? | The root disk and every volume attached when the backup runs. You can't leave a volume out. | | What's not included? | Volumes attached after the run, or detached before it. Bare metal servers. | | When does it run? | **Daily** or **Weekly**, at an hour you choose, in your browser's time zone. Each run starts within 30 minutes after that hour. The default is **Daily** at **12:00**. | | How long are recovery points kept? | **7 days**. They are deleted a weekly chain at a time, once the newest point in the chain is more than 7 days old, so you can see up to about 14 days of points. The newest chain is never deleted automatically. | | What does it cost? | Backup storage per GB per month, for as long as each recovery point exists. See [Backup pricing](/docs/tools/backups/backup-pricing). | | Is it consistent? | It's **crash-consistent**: like pulling the power cord. Files written before the run are there; writes in progress may be incomplete. | | Does a restore overwrite my VM? | **Replace Current VM** does: everything written since the recovery point is lost. **Create New VM** and **Restore Selected Disk** leave the VM unchanged. | | Can I restore in another location? | Not from the portal. A new VM from a backup is created in the original VM's location. | | What happens when I delete the VM? | The schedule stops. Existing recovery points are kept, and billed, until you delete them on **Tools → Backups**. | ## Turn on backups ### Open the VM's Backups tab Open the VM (**Infrastructure → Cloud VMs** or **GPU VMs**) and click **Backups**. ### Turn on Automated Backups Switch **Automated Backups** to **Enabled**. The portal confirms "Backup enabled with default schedule (daily 12:00 …)". ### Change the schedule if needed Pick **Daily** or **Weekly** under **Schedule backups**, the day under **On** (weekly only) and the hour under **At**, then click **Save Schedule**. See [Enable backups](/docs/tools/backups/enable-backups) for every option. ## Where to find backups | Place | Use it for | | ------------------------ | ------------------------------------------------------------------------------------------------------------- | | The VM's **Backups** tab | Turn backups on or off, set the schedule, see the VM's **Recovery Points**, open **Details**, and **Restore** | | **Tools → Backups** | All backups in the workspace. Each row's menu has **Convert to Snapshot** and **Delete Backup**. | The VM's **Recovery Points** list shows: | Column | Meaning | | ------------------ | ---------------------------------------------------------------------------------------- | | **Recovery Point** | When it was taken, and its ID | | **Topology** | The disks captured, for example **Root + 1 data disk** | | **Drift** | **In sync**, or **Drifted** when the VM's volumes have changed since the point was taken | | **Status** | Whether the run is queued, running, succeeded, failed or cancelled | | **Created** | Date and time | | **Storage** | The size of the recovery point | Until the first backup runs, the list says **No recovery points yet**. ## Restore options **Restore** is available on successful recovery points: | Mode | What happens | | ------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Replace Current VM** | The VM is stopped, its root disk is replaced, and the captured data disks are recreated in their saved order. Changes since the recovery point are lost. | | **Create New VM** | A new VM, with new volumes, is built from the recovery point. The original VM isn't touched. | | **Restore Selected Disk** | One captured disk becomes a new, detached volume that you can attach later. The VM isn't changed. | See [Restore from backup](/docs/tools/backups/restore-from-backup). ## Remote backup in another data centre If you need copies kept outside the data centre where your VM runs, IBEE offers remote backup to a separate data centre as a managed service. It isn't turned on from the portal: the IBEE team sets it up on request and shares the details and pricing for your setup. Contact [sales@ibee.ai](mailto:sales@ibee.ai) to arrange it. ## Use the API or CLI Per-VM backups are in the public API. GPU VMs use the same paths under `/compute/gpu-vms/{vm_id}` and `/compute/gpu-vm-backups`, and the CLI's `ibee gpus` group. | Task | API | CLI | | --------------------------------- | ----------------------------------------------------------------------- | -------------------------------------------------- | | Read or change the schedule | `GET` / `PATCH /compute/cloud-vms/{vm_id}/backups/policy` | `ibee vms backup-policy get` / `update` | | Turn backups on or off | `POST /compute/cloud-vms/{vm_id}/backups/enable`, `.../backups/disable` | `ibee vms backup-policy enable` / `disable` | | List a VM's recovery points | `GET /compute/cloud-vms/{vm_id}/backups/runs` | `ibee vms backups list VM_ID` | | Take a backup now | `POST /compute/cloud-vms/{vm_id}/backups/runs` | `ibee vms backups create VM_ID` | | Restore | `POST /compute/cloud-vms/{vm_id}/backups/actions/restore` | `ibee vms backups restore VM_ID RECOVERY_POINT_ID` | | List all backups in the workspace | `GET /compute/cloud-vm-backups/runs` | | | Delete a backup | `DELETE /compute/cloud-vm-backups/runs/{run_id}` | | The two list calls return only finished, successful runs. To see a queued, running or failed run, read it by ID with `GET /compute/cloud-vm-backups/runs/{run_id}`. For a scheduled run, the `last_status` field of the backup policy shows how the latest finished run ended. Enabling backups and taking one now need the backup storage price entry as `billing_catalog`. Get it from `GET /compute/plans/backup-storage?workspace_id=...` and copy a `selectable` item's `billing_catalog`. Through the API, the schedule can also be `hourly`, and `retention_days` can be 1 to 365. **Convert to Snapshot** is portal-only. ## Troubleshooting | You see | What it means | What to do | | -------------------------------------------------------------------------------------------------- | -------------------------------------------------------- | ----------------------------------------------------------------------------------------------------- | | **No Backups** on **Tools → Backups** | No VM in the workspace has backups turned on | Click **Manage Servers**, open a VM and turn on **Automated Backups**. | | **No recovery points yet** | Backups are on, but the first scheduled time hasn't come | Wait for the scheduled hour (plus up to 30 minutes). | | "Backup cannot be deleted because a newer recovery point depends on its incremental chain" (`409`) | Later incremental points need this one | Delete the newer points in the chain first, or wait for the chain to expire. | | "Backup cannot be deleted while a restore is still running" | A restore is using this recovery point | Wait for the restore to finish. | | "Only a completed backup can be deleted" | The backup is still `queued` or `running` | Wait until it finishes. | | "Backup run already queued or running for this VM" (`409`) | A backup is already in progress | Wait for it to finish before starting another. | | A recovery point shows **Drifted** | Volumes were attached or detached since it was taken | It still restores what it captured. Check **Details** before you restore. | | A status is `failed` | That run didn't complete | Check [Activity logs](/docs/platform-fundamentals/activity-logs). The next scheduled run tries again. | ## Related pages * [Enable backups](/docs/tools/backups/enable-backups) * [Restore from backup](/docs/tools/backups/restore-from-backup) * [Backup pricing](/docs/tools/backups/backup-pricing) * [Snapshots](/docs/tools/snapshots) ## Docs - [Enable backups](https://ibee.ai/docs/tools/backups/enable-backups.md): Turn on automatic daily or weekly backups for a Cloud VM or GPU VM, choose the time, check that the first backup ran, and turn backups off. With the portal, CLI and API. - [Restore from backup](https://ibee.ai/docs/tools/backups/restore-from-backup.md): Recover a Cloud VM or GPU VM from an automatic backup. Choose between replacing the VM, creating a new VM, or restoring one disk; what each mode overwrites; and how to mount restored volumes. - [Backup pricing](https://ibee.ai/docs/tools/backups/backup-pricing.md): How VM backups are billed. You pay per GB per month for the storage recovery points use, for as long as they exist; running a backup or a restore costs nothing extra. How to find the price, estimate a bill and keep it down.