Backup

The most important thing to back up is the Postgres database, that’s where all user data lives. The media volume can usually be re-fetched from upstream, and static files are regenerated on every container start.

You should perform a test restore at least once, so you know the procedure works when you actually need it.

Database

Make a dump of the database:

# Stop the other services so the database isn't changed mid-export
docker compose stop web nginx cache celery_worker celery_beat
docker compose exec db pg_dumpall --clean --username wger > backup.sql
docker compose start

To restore from a dump:

docker compose stop
docker volume remove docker_postgres-data
docker compose up db
cat backup.sql | docker compose exec -T db psql --username wger --dbname wger
docker compose up

The PowerSync bucket-storage tables (powersync.* schema) are effectively a cache rebuilt from the public.* wger tables and sync_rules.yaml. Nothing in there is a primary source of truth, every row can be regenerated via a fresh snapshot against the main wger database.

pg_dumpall above captures the schema automatically, so there is nothing extra to do. If backup size matters (bucket_data scales with users × sync-rule complexity), you can safely exclude it:

docker compose exec db pg_dump --username wger \
    --exclude-schema=powersync --clean --create wger > backup.sql

On restore, PowerSync re-bootstraps the schema on next startup and takes a fresh snapshot. Clients silently re-sync their local SQLite from scratch.

Important

After restoring the database, drop the PowerSync replication slot and restart the service so it re-reads from a valid WAL position:

docker compose exec db psql -U wger -c \
    "SELECT pg_drop_replication_slot(slot_name) \
     FROM pg_replication_slots \
     WHERE slot_name LIKE 'powersync_%';"
docker compose restart powersync

Skipping this leaves the slot pointing at a WAL position that no longer exists, and PowerSync fails with cryptic operator does not exist errors until it is re-created.

Media

If you haven’t uploaded your own exercise images, exercise videos or gallery images, you don’t need to back up the media volume, the contents can be re-downloaded from the upstream wger instance. Truncate the relevant tables and run the sync commands again:

docker compose exec db psql -U wger -c "TRUNCATE TABLE exercises_exerciseimage, exercises_exercisevideo;"
docker compose exec db psql -U wger -c "TRUNCATE TABLE nutrition_image;"

docker compose exec web python3 manage.py download-exercise-images
docker compose exec web python3 manage.py download-exercise-videos

If you do have uploaded media that needs to be preserved, such as gallery entries, consult these options for backing up Docker volumes:

Static files

The contents of the static volume are 100% generated and recreated on startup, no need to back up anything.