Backups

What Gets Backed Up

PostgreSQL (notes, templates, collaboration metadata)
Redis keyspace (sessions, rate-limiter buckets, share-link cache)
Generated archives are encrypted and streamed to the configured S3-compatible bucket.

The backup runner is controlled by environment variables outlined in /getting-started/environment-variables.

Configuration

Set these variables before launching the backup service:

ENABLE_BACKUPS=true
BACKUP_S3_ENDPOINT=https://s3.your-cloud.com
BACKUP_S3_BUCKET=leaflock-backups
BACKUP_S3_ACCESS_KEY=...
BACKUP_S3_SECRET_KEY=...
BACKUP_SCHEDULE="0 */6 * * *"   # every six hours
BACKUP_ENCRYPTION_KEY=<32-byte base64>

The schedule uses cron syntax. Choose a frequency that matches your RPO/RTO requirements.

Observability

Metrics exposed at /metrics:

leaflock_backups_total{status="success|failed"} – run outcomes.
leaflock_backup_duration_seconds – histogram of runtime.
leaflock_backup_size_bytes – last archive size (compressed).

Restore Test (Docker Example)

# Stop write traffic
docker compose stop backend frontend

# Restore PostgreSQL
aws s3 cp s3://leaflock-backups/latest.sql.gz - | \
gunzip | docker compose exec -T postgres psql -U postgres notes

# Restore Redis (if snapshotted)
aws s3 cp s3://leaflock-backups/latest.rdb ./dump.rdb
docker compose cp dump.rdb redis:/data/dump.rdb
docker compose start backend frontend

Adapt commands for Kubernetes by using kubectl exec or job-based restores.

Failure Modes

No archives uploading: check runner logs for S3 credential errors. Ensure network access to the endpoint.
Backups completing but too large: verify garbage collection of deleted notes; run cleanup_old_deleted_notes() via services/cleanup.go.
Restore missing templates: include the template seed files (backend/templates/) in source control; backups capture database rows, not local seed files.

Document each successful restore drill to satisfy compliance and keep muscle memory fresh.