Skip to content

Arkive user guide

Arkive backs up the Docker apps on an Unraid server. It finds the databases behind your containers, dumps each one with its own tool, and stores the dumps, your Unraid flash configuration and the folders you choose as encrypted restic snapshots on storage you control. After that it keeps checking that backups are still happening and that the dumps actually restore.

Arkive dashboard: the "Finish protecting this server" checklist with four of five steps done, then system health, last backup, storage used and the backup trust score

Where to start

If you want to Read
Install Arkive on Unraid Quick start on Unraid
Understand each step of first-time setup Setup wizard
Know what is left before the server counts as protected Finish protecting this server
Choose and configure where backups go Storage providers
Get alerts when something goes wrong Notifications and Overdue backups and heartbeat
Make sure you can recover if the server dies Recovery kit
Get files or a database back Restoring backups, including the guided database restore
Prove the database dumps restore Restore drills
See each database's last dump, size trend and last drill Database health
Find out why a run is partial or a job is overdue Troubleshooting
Move to a newer version Upgrading
Decide how to expose Arkive on your network Security and deployment model

What a backup run does

Each backup job runs the same pipeline:

  1. Discover. For jobs that dump databases, Arkive lists your containers through the Docker socket and finds their databases: PostgreSQL, MariaDB/MySQL, MongoDB, SQLite and Redis.
  2. Dump. Each database is dumped with its native tool into /config/dumps (Full Backup and Database Dumps jobs). A dump much smaller than that database's recent dumps makes the run partial (see Database health).
  3. Archive the flash drive. On Unraid, /boot/config is archived into the same folder as flash_<timestamp>.tar (Full Backup and Flash Drive jobs).
  4. Snapshot. restic backs up the dump folder and every enabled watched directory to each storage target of the job. A copy of Arkive's own database (arkive_self.db) goes into the dump folder first, so it is in every snapshot too.
  5. Apply retention. restic forget applies your keep-daily, keep-weekly and keep-monthly settings on each target. The space is reclaimed by a prune that runs every Sunday at 04:00.
  6. Report. The run ends as success, partial (something was missed, with the reason) or failed, and a notification goes to the channels that subscribed to that outcome.

Because every job uploads the same set of paths, every snapshot is a complete restore point. A partial run usually still wrote a snapshot, but not always: a job with no storage target to upload to ends as partial with "nothing was uploaded" and leaves no restore point at all.

Words used in this guide

  • Job: a scheduled backup. There are three types. Full Backup refreshes the database dumps and the flash archive, Database Dumps refreshes only the dumps, and Flash Drive refreshes only the flash archive. All three then upload a complete snapshot.
  • Target: a storage destination (a local folder, SFTP server, B2, S3-compatible bucket, Wasabi, Dropbox or Google Drive). Each target holds its own restic repository.
  • Snapshot: one restic backup in a target's repository.
  • Run status: success, partial, failed, skipped (a scheduled run that could not start because another backup or restore was still running after 3 hours) or cancelled.

Try the interface with demo data

Add ?demo=true to any Arkive page, for example http://your-server:8200/?demo=true. The web interface then uses sample data built into the frontend (containers, databases, snapshots, restore drills, the setup checklist) instead of your server's API, so you can look around without signing in. Nothing you do in demo mode changes your configuration. Reload the page without the parameter to leave demo mode.

The screenshots in this guide and in the README were taken in demo mode. frontend/scripts/docs-screenshots.sh takes them again after an interface change.