Introduction
This project provides Docker images to periodically back up a PostgreSQL database to AWS S3, and to restore from the backup as needed.
Usage
Backup
services: postgres: image: postgres:16 environment: POSTGRES_USER: user POSTGRES_PASSWORD: password backup: image: eeshugerman/postgres-backup-s3:16 environment: SCHEDULE: '@weekly' # optional BACKUP_KEEP_DAYS: 7 # optional PASSPHRASE: passphrase # optional S3_REGION: region S3_ACCESS_KEY_ID: key S3_SECRET_ACCESS_KEY: secret S3_BUCKET: my-bucket S3_PREFIX: backup POSTGRES_HOST: postgres POSTGRES_DATABASE: dbname POSTGRES_USER: user POSTGRES_PASSWORD: password
- Images are tagged by the major PostgreSQL version supported:
12,13,14,15or16. - The
SCHEDULEvariable determines backup frequency. See go-cron schedules documentation here. Omit to run the backup immediately and then exit. - If
PASSPHRASEis provided, the backup will be encrypted using GPG. - Run
docker exec <container name> sh backup.shto trigger a backup ad-hoc. - If
BACKUP_KEEP_DAYSis set, backups older than this many days will be deleted from S3. - Set
S3_ENDPOINTif you're using a non-AWS S3-compatible storage provider.
Restore
Caution
DATA LOSS! All database objects will be dropped and re-created.
... from latest backup
docker exec <container name> sh restore.sh
Note
If your bucket has more than a 1000 files, the latest may not be restored -- only one S3 ls command is used
... from specific backup
docker exec <container name> sh restore.sh <timestamp>
Development
Build the image locally
ALPINE_VERSION determines Postgres version compatibility. See build-and-push-images.yml for the latest mapping.
DOCKER_BUILDKIT=1 docker build --build-arg ALPINE_VERSION=3.14 .Run a simple test environment with Docker Compose
cp template.env .env
# fill out your secrets/params in .env
docker compose up -dAcknowledgements
This project is a fork and re-structuring of @schickling's postgres-backup-s3 and postgres-restore-s3.
Fork goals
These changes would have been difficult or impossible merge into @schickling's repo or similarly-structured forks.
- dedicated repository
- automated builds
- support multiple PostgreSQL versions
- backup and restore with one image
Other changes and features
- some environment variables renamed or removed
- uses
pg_dump'scustomformat (see docs) - drop and re-create all database objects on restore
- backup blobs and all schemas by default
- no Python 2 dependencies
- filter backups on S3 by database name
- support encrypted (password-protected) backups
- support for restoring from a specific backup by timestamp
- support for auto-removal of old backups