> For the complete documentation index, see [llms.txt](https://docs.ellipsis-drive.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.ellipsis-drive.com/installation/system-architecture-and-configuration/backup-and-recovery.md).

# Backup and recovery

#### User-level recovery (soft delete model)

When a user deletes data in Ellipsis Drive, the system applies a **soft delete** followed by a **delayed hard delete**.

This ensures that accidental deletions can be recovered within a safe recovery window.

Users can restore deleted data directly from their personal trash via the Ellipsis Drive interface.

If a user permanently empties the trash, data can still be restored via the system dashboard for a limited period.

After approximately **2 months**, data is permanently removed from the system.

***

#### System-level recovery

In cases of full deletion or system-level failure, data can be restored using system backups.

***

#### Ellipsis Drive-hosted Private Instance

In a hosted Private Instance, backup and recovery are fully managed by Ellipsis Drive.

Key characteristics:

* Daily backups are performed automatically
* Backups are retained for 4 months
* Backups are stored across two geographically separate locations
* Recovery typically takes 2–3 hours depending on dataset size and system state

***

#### Self-hosted Private Instance

If you operate your own Private Instance, you are responsible for all backup and recovery operations.

The following components must be included in your backup strategy:

* File storage node
* Central metadata storage
* Raster activation storage pool
* Vector activation storage pool

While raster and vector activation pools can technically be reconstructed from file storage, backups are strongly recommended to ensure faster recovery times and reduced operational risk.

***

#### Recovery procedure

To restore the system to a previous state:

1. Replace the disks of all relevant nodes with the backup versions
2. Restart all API instances
3. Clear or refresh cache layers to ensure consistency across services

***