Back up Assure

Use this procedure on the Docker host. You need access to the database and data volumes.

Back up both volumes together: pg (the database) and data (signing keys and checkpoints). Without the signing key, new evidence packs get a new fingerprint and you'll need to send it to your auditor again. Pause feeds meanwhile, because it writes signed checkpoints of the ledger, so the two copies match. Write the backups outside the checkout (deploy/update.sh refuses untracked files in it):

docker compose stop feeds

docker compose exec -T db pg_dump -U nimstead nimstead > ~/nimstead.sql

docker compose cp app:/data ~/nimstead-data

docker compose start feeds

Run this at least daily and keep copies off the host. What a restore loses is what changed since the last backup: inventories come back at the agents' next 6-hourly run, but ledger entries made after the backup are gone. With checkpoints in object-locked S3, a restore from before the newest checkpoint shows as a shortened ledger on Evidence. That's the tamper evidence doing its job: restore the newest backup, and note the restore in your records for the auditor.

Check that both backup outputs exist and are readable by the operator who will restore them. Protect the data backup as a secret because it contains signing keys. Test the restore procedure on an isolated host.