django-rclone

Django database and media backup management commands, powered by rclone.

django-rclone bridges Django's database layer with rclone's file transfer layer. You get native database dumps piped directly to any of rclone's 70+ supported cloud storage backends -- no temp files, no intermediate archives, no Python reimplementations of what rclone already does.

Full documentation

Why rclone instead of Django Storages?

django-dbbackup is a mature and well-regarded backup solution. It wraps Django Storages for upload, implements GPG encryption in Python, handles gzip compression, and parses filenames with regex to manage backups.

django-rclone takes a different approach: delegate everything that isn't Django-specific to rclone.

Concern	django-dbbackup	django-rclone
Storage backends	Django Storages (S3, GCS, etc.)	rclone (70+ backends natively)
Encryption	GPG subprocess wrapper in Python	rclone crypt remote
Compression	gzip in Python	rclone compress remote or --compress flag
Media backup	Tar archive, then upload	rclone sync (incremental, no archiving)
Backup listing	Filename regex parsing	rclone lsjson (structured JSON)
Temp files	SpooledTemporaryFile	None -- pipes directly via rclone rcat

The result is significantly less code doing significantly less work. Storage abstraction, encryption, compression, and incremental sync are all rclone's problem -- django-rclone only owns what Django must own: database connectors, management commands, and signals.

Requirements

• Python 3.12+
• Django 5.2+
• rclone installed and configured

Installation

pip install django-rclone

Add to your INSTALLED_APPS:

INSTALLED_APPS = [
    # ...
    "django_rclone",
]

Configure your rclone remote (see rclone docs):

rclone config

Then point django-rclone at it:

DJANGO_RCLONE = {
    "REMOTE": "myremote:backups",
}

Usage

Database backup and restore

# Backup the default database
python manage.py dbbackup

# Backup a specific database
python manage.py dbbackup --database analytics

# Backup and clean old backups beyond retention count
python manage.py dbbackup --clean

# Restore from the latest backup
python manage.py dbrestore

# Restore a specific backup
python manage.py dbrestore --input-path default-2024-01-15-120000.dump

# Non-interactive restore (for automation)
python manage.py dbrestore --noinput --input-path default-2024-01-15-120000.dump

Media backup and restore

# Sync MEDIA_ROOT to remote (incremental -- only changed files transfer)
python manage.py mediabackup

# Sync remote back to MEDIA_ROOT
python manage.py mediarestore

List backups

# List all database backups
python manage.py listbackups

# Filter by database
python manage.py listbackups --database default

# List media files on remote
python manage.py listbackups --media

Configuration

All settings live under the DJANGO_RCLONE dict in your Django settings:

DJANGO_RCLONE = {
    # Required -- rclone remote and base path
    "REMOTE": "myremote:backups",

    # Optional -- rclone binary and config
    "RCLONE_BINARY": "rclone",             # Path to rclone binary
    "RCLONE_CONFIG": None,                 # Path to rclone.conf (None uses default)
    "RCLONE_FLAGS": [],                    # Extra flags for every rclone call

    # Database backup settings
    "DB_BACKUP_DIR": "db",                 # Subdirectory for DB backups
    "DB_FILENAME_TEMPLATE": "{database}-{datetime}.{ext}",  # Must start with {database}
    "DB_DATE_FORMAT": "%Y-%m-%d-%H%M%S",
    "DB_CLEANUP_KEEP": 10,                 # Keep N most recent backups per database

    # Media backup settings
    "MEDIA_BACKUP_DIR": "media",           # Subdirectory for media backups

    # Database connector overrides
    "CONNECTORS": {},                      # Per-database connector class overrides
    "CONNECTOR_MAPPING": {},               # Engine-to-connector class overrides
}

Encryption and compression

django-rclone does not implement encryption or compression. Instead, configure these at the rclone level where they belong:

Encryption -- use a crypt remote:

rclone config create myremote-crypt crypt remote=myremote:backups password=your-password

Then set "REMOTE": "myremote-crypt:" in your Django settings.

Compression -- use a compress remote:

rclone config create myremote-compressed compress remote=myremote:backups

Or pass --compress-level via RCLONE_FLAGS.

See Storage Providers for provider-specific configuration notes (Cloudflare R2, etc.).

Supported databases

Database	Connector	Dump tool	Format
PostgreSQL	PgDumpConnector	pg_dump / pg_restore	Custom (binary)
PostGIS	PgDumpGisConnector	pg_dump / pg_restore	Custom (binary)
MySQL / MariaDB	MysqlDumpConnector	mysqldump / mysql	SQL text
SQLite	SqliteConnector	sqlite3 .dump	SQL text
MongoDB	MongoDumpConnector	mongodump / mongorestore	Archive (binary)

GIS backends (postgis, spatialite, gis/mysql) and django-prometheus wrappers are also mapped automatically. See connectors documentation for the full engine mapping table.

Signals

django-rclone sends Django signals before and after each operation:

from django_rclone.signals import pre_db_backup, post_db_backup

@receiver(post_db_backup)
def notify_on_backup(sender, database, path, **kwargs):
    logger.info("Database %s backed up to %s", database, path)

Available signals: pre_db_backup, post_db_backup, pre_db_restore, post_db_restore, pre_media_backup, post_media_backup, pre_media_restore, post_media_restore.

Architecture

Management Commands (dbbackup, dbrestore, mediabackup, mediarestore, listbackups)
        |                           |
   DB Connectors              rclone.py
   (pg, mysql, sqlite,    (subprocess wrapper)
    mongodb)
        |                           |
   Database binary              rclone binary
                          (70+ storage backends)

Database dumps stream directly from the dump process into rclone rcat via Unix pipes. No intermediate files are written. Restores work in reverse: rclone cat streams into the database restore process. Subprocess finalization is centralized and deadlock-safe (pipe draining + stderr collection).

Media backups use rclone sync, which is incremental by default -- only changed files are transferred.

Development

Contributions are welcome. This project enforces 100% test coverage -- all new code must be fully covered by tests. The CI pipeline will fail if coverage drops below 100%.

CI also includes subprocess guardrail tests to prevent wait()-based pipe deadlocks and to keep raw Popen(...) usage confined to the wrapper modules.

uv sync                                  # Install runtime + dev deps (tests/lint/type/docs)
uv run pytest --cov --cov-branch         # Unit tests (integration excluded by default)
uv sync --group integration               # Install DB integration test dependencies
uv run pytest tests/integration -m integration -q       # Integration tests (requires tools/services)
uv run ruff check .                      # Lint
uv run ruff format --check .             # Check formatting
uv run ty check                          # Type check
uv run mkdocs build --strict             # Build docs

For full integration setup details (Docker services, required binaries, env vars), see docs/testing.md.

License

MIT