Consistent partial database dump utility

database, dump, postgresql, cli, django
pip install xdump==0.6.0



Build Status Coverage Status Documentation Status Latest PyPI version

XDump is a utility to make a consistent partial dump and load it into the database.

The idea is to provide an ability to specify what to include in the dump via SQL queries.


XDump can be obtained with pip:

$ pip install xdump

Usage example

Make a dump (on production replica for example):

>>> from xdump.postgresql import PostgreSQLBackend
>>> backend = PostgreSQLBackend(dbname='app_db', user='prod', password='pass', host='', port='5432')
>>> backend.dump(
    partial_tables={'employees': 'SELECT * FROM employees ORDER BY id DESC LIMIT 2'}

Load a dump on your local machine:

>>> backend = PostgreSQLBackend(dbname='app_db', user='local', password='pass', host='', port='5432')
# If you need a clear DB
>>> backend.recreate_database()  # or `backend.truncate()`
>>> backend.load('/path/to/')

Dump is compressed by default. Compression level could be changed with passing compression argument to dump method. Valid options are zipfile.ZIP_STORED, zipfile.ZIP_DEFLATED, zipfile.ZIP_BZIP2 and zipfile.ZIP_LZMA.

The verbosity of the output could be customized via verbosity (with values 0, 1 or 2) argument of a backend class.

There are two options to control the content of the dump:

  • dump_schema - controls if the schema should be included
  • dump_data - controls if the data should be included

Automatic selection of related objects

You don't have to specify all queries for related objects - XDump will load them for you automatically. It covers both, recursive and non-recursive relations. For example, if the employees table has foreign keys group_id (to groups table) and manager_id (to employees table) the resulting dump will have all objects related to selected employees (as well as for objects related to related objects, recursively).

Command Line Interface

xload provides an ability to create a dump.


xdump [postgres|sqlite] [OPTIONS]

Common options:

-o, --output TEXT               output file name  [required]
-f, --full TEXT                 table name to be fully dumped. Could be used
                                multiple times
-p, --partial TEXT              partial tables specification in a form
                                "table_name:select SQL". Could be used
                                multiple times
-c, --compression [deflated|stored|bzip2|lzma]
                                dump compression level
--schema / --no-schema          include / exclude the schema from the dump
--data / --no-data              include / exclude the data from the dump
-D, --dbname TEXT               database to work with  [required]
-v, --verbosity                 verbosity level

PostgreSQL-specific options:

-U, --user TEXT                 connect as specified database user
-W, --password TEXT             password for the DB connection
-H, --host TEXT                 database server host or socket directory
-P, --port TEXT                 database server port number

xload loads a dump into a database.


xload [postgres|sqlite] [OPTIONS]

Common options:

-i, --input TEXT                input file name  [required]
-m, --cleanup-method [recreate|truncate]
                                method of DB cleaning up
-D, --dbname TEXT               database to work with  [required]
-v, --verbosity                 verbosity level

PostgreSQL-specific options are the same as for xdump.

RDBMS support

At the moment only the following are supported:

  • PostgreSQL
  • SQLite >= 3.8.3

Django support

Add xdump.extra.django to your INSTALLED_APPS settings:


Add XDUMP to your project settings file. It should contain minimum two entries:

  • FULL_TABLES - a list of tables that should be fully dumped.
  • PARTIAL_TABLES - a dictionary with table_name: select SQL
    'FULL_TABLES': ['groups'],
    'PARTIAL_TABLES': {'employees': 'SELECT * FROM employees WHERE id > 100'}

Optionally you could use a custom backend:

    'BACKEND': 'importable.string',

Run xdump command:

$ ./ xdump

Run xload command:

$ ./ xload

Possible options to both commands:

  • -a/--alias - allows you to choose database config from DATABASES, that is used during the execution;
  • -b/--backend - importable string, that leads to custom dump backend class.

Options for xdump command:

  • -s/--dump-schema - controls if the schema should be included;
  • -d/--dump-data - controls if the data should be included.

Options for xload command:

  • -m/--cleanup-method - optionally re-creates DB or truncates the data.

NOTE. If the dump has no schema inside, DB won't be re-created.

The following make command could be useful to get a configured dump from production to your local machine:

    ssh -t $(TARGET) "DJANGO_SETTINGS_MODULE=settings.production /path/to/ xdump /tmp/"
    scp $(TARGET):/tmp/ ./
    ssh -t $(TARGET) "rm /tmp/"
    DJANGO_SETTINGS_MODULE=settings.local $(PYTHON) xload ./

And the usage is:

$ make sync-production PYTHON=/path/to/python/in/venv

Python support

XDump supports Python 2.7, 3.4 - 3.7 and PyPy 2 & 3.