Self-hosting

Backups

Back up and restore a self-hosted Notifuse: every database, and the secret key.

Notifuse keeps its data in several PostgreSQL databases: notifuse_system (users, workspaces, settings) and one notifuse_ws_<workspace id> database per workspace (contacts, lists, templates, broadcasts, message history, web analytics). A backup needs all of them, and the SECRET_KEY the installation runs with: the stored provider credentials are encrypted with it, and a restore without it leaves them unreadable.

Warning

A backup of notifuse_system alone restores an installation that looks healthy, with its users and workspaces, and holds no contact, broadcast or message. Back up every notifuse_ws_… database with it.

Backing up

With the provided compose.yaml, from the directory that holds it (replace postgres with your DB_USER, and notifuse with your DB_PREFIX, if you changed them):

for db in $(docker compose exec -T postgres psql -U postgres -d postgres -Atc \
    "SELECT datname FROM pg_database WHERE datname = 'notifuse_system' OR datname LIKE 'notifuse\_ws\_%'"); do
  docker compose exec -T postgres pg_dump -U postgres -Fc "$db" > "$db.dump"
done

Each database lands in its own .dump file. Copy them, with SECRET_KEY, somewhere other than the server. The dumps are consistent per database; for a backup consistent across them, stop Notifuse first (docker compose stop api) and start it again after (docker compose start api).

pg_dumpall is not a good fit: its restore drops and recreates the role it connects as, which fails, and it writes the role’s password hash into the backup.

Copying the files of the database volume while PostgreSQL runs does not give a restorable backup.

Restoring

Into an installation whose database is empty (a new server, or after docker compose down -v), with the same SECRET_KEY in .env and DB_EXPECT_EXISTING unset:

docker compose up -d postgres
for f in notifuse_*.dump; do
  db=${f%.dump}
  docker compose exec -T postgres createdb -U postgres "$db"
  docker compose exec -T postgres pg_restore -U postgres -d "$db" --exit-on-error < "$f"
done
docker compose up -d

Restore before Notifuse starts on the empty database: a first start creates a new, empty notifuse_system. Once the restore works, set DB_EXPECT_EXISTING=true in .env, so that Notifuse refuses to start the day it finds an empty database instead of your data.