🚍 💜 👱🏿 PgGraph - utility for archiving and finding table dependencies in PostgreSQL 🗄️ 🦃 😘

Today I want to present to the readers of Habr a utility written in Python for working with table dependencies in the PostgreSQL DBMS.

The utility API is simple and consists of three methods:

archive_table - recursive archiving / deleting rows with specified Primary Keys
get_table_references - find dependencies for a table (will show tables referenced and referenced by the specified one)
get_rows_references - search for rows in other tables that refer to the specified rows in the desired table

Background

My name is Oleg Borzov, I am a developer on the CRM team for mortgage loan managers at Domklik.

The main database of our CRM system is one of the largest in terms of volume in the company. It is one of the oldest: it appeared at the very launch of the project, when the trees were large, Domclick was a startup, and instead of a microservice on a fashionable Python asynchronous framework, there was a huge monolith in PHP.

The transition from PHP to Python was very long and required the simultaneous support of both systems, which affected the design of the database.

As a result, we have a database with a large number of highly connected and huge tables with a bunch of indexes for different types of queries. All this negatively affects the performance of the database: because of large tables and a heap of connections between them, the complexity of queries is constantly growing, which is especially critical for the most loaded tables.

To reduce the load on the database, we decided to write a script that would transfer old records from the most voluminous and loaded tables to archive tables (for example, from taskc task_archive) every day .

This task is complicated by a large number of relationships between tables: just transferring rows from taskto is task_archivenot enough, before that you need to do the same recursively with all referring tasktables.

I'll demonstrate with an exampledemo database from postgrespro.ru :

Let's say we need to delete records from a table Flights. Postgres will not allow us to do this just like that: we first need to delete records from all referring tables, and so recursively to tables that no one refers to.

In our example, it Flightsrefers to Ticket_flights, and to it - Boarding_passes.

Therefore, you need to delete in this order:

Get the values of the Primary Keys (PK) of rows in Ticket_flights, which refer to the deleted rows in Flights.
We get the PK of the lines Boarding_passesthat refer to Ticket_flights.
We delete lines by PK from item 2 in the table Boarding_passes.
We delete lines by PK from clause 1 in Ticket_flights.
We delete lines from Flights.

The result is a utility called PgGraph, which we decided to make open source.

How to use

The utility supports two modes of use:

Call from the command line ( pggraph …).
Usage in Python code (class PgGraphApi).

Installation and configuration

First you need to install the utility from the Pypi repository:

pip3 install pggraph

Then create a config.ini file on the local machine with the database configuration and the archive script:

[db]
host = localhost
port = 5432
user = postgres
password = postgres
dbname = postgres
schema = public ;  ,    

[archive]  ;    ,     
is_debug = false
chunk_size = 1000
max_depth = 20
to_archive = true
archive_suffix = 'archive'

Running from the console

Parameters

$ pggraph -h
usage: pggraph action [-h] --table TABLE [--ids IDS] [--config_path CONFIG_PATH]
positional arguments:
  action        required action: archive_table, get_table_references, get_rows_references

optional arguments:
  -h, --help                    show this help message and exit
  --table TABLE                 table name
  --ids IDS                     primary key ids, separated by comma, e.g. 1,2,3
  --config_path CONFIG_PATH     path to config.ini
  --log_path LOG_PATH           path to log dir
  --log_level LOG_LEVEL         log level (debug, info, error)

Positional arguments:

action- the desired effect: archive_table, get_table_referencesor get_rows_references.

Named arguments:

--config_path - path to the config file;
--table - the table with which you want to perform the action;
--ids- list of id separated by commas, for example, 1,2,3(optional);
--log_path - path to the folder for logs (optional parameter, by default - home folder);
--log_level - logging level (optional parameter, by default - INFO).

Command examples

Archiving a table

The main function of the utility is data archiving, i.e. transferring rows from the main table to the archive table (for example, from the books table to books_archive ).

Deletion without archiving is also supported: to do this, set the parameter to_archive = false in config.ini ).

Required parameters are config_path, table and ids .

After startup, records idsin the table tableand in all tables referring to it will be recursively deleted .

$ pggraph archive_table --config_path config.hw.local.ini --table flights --ids 1,2,3
2020-06-20 19:27:44 INFO: flights - START
2020-06-20 19:27:44 INFO: flights - start archive_recursive 3 rows (depth=0)
2020-06-20 19:27:44 INFO:       START ARCHIVE REFERRING TABLES
2020-06-20 19:27:44 INFO:       ticket_flights - start archive_recursive 3 rows (depth=1)
2020-06-20 19:27:44 INFO:               START ARCHIVE REFERRING TABLES
2020-06-20 19:27:44 INFO:               boarding_passes - start archive_recursive 3 rows (depth=2)
2020-06-20 19:27:44 INFO:                       START ARCHIVE REFERRING TABLES
2020-06-20 19:27:44 INFO:                       END ARCHIVE REFERRING TABLES
2020-06-20 19:27:44 INFO:               boarding_passes - archive_by_ids 3 rows by ticket_no, flight_id
2020-06-20 19:27:44 INFO:               boarding_passes - start archive_recursive 3 rows (depth=2)
2020-06-20 19:27:44 INFO:                       START ARCHIVE REFERRING TABLES
2020-06-20 19:27:44 INFO:                       END ARCHIVE REFERRING TABLES
2020-06-20 19:27:44 INFO:               boarding_passes - archive_by_ids 3 rows by ticket_no, flight_id
2020-06-20 19:27:44 INFO:               boarding_passes - start archive_recursive 3 rows (depth=2)
2020-06-20 19:27:44 INFO:                       START ARCHIVE REFERRING TABLES
2020-06-20 19:27:44 INFO:                       END ARCHIVE REFERRING TABLES
2020-06-20 19:27:44 INFO:               boarding_passes - archive_by_ids 3 rows by ticket_no, flight_id
2020-06-20 19:27:44 INFO:               boarding_passes - start archive_recursive 3 rows (depth=2)
2020-06-20 19:27:44 INFO:                       START ARCHIVE REFERRING TABLES
2020-06-20 19:27:44 INFO:                       END ARCHIVE REFERRING TABLES
2020-06-20 19:27:44 INFO:               boarding_passes - archive_by_ids 3 rows by ticket_no, flight_id
2020-06-20 19:27:44 INFO:               END ARCHIVE REFERRING TABLES
2020-06-20 19:27:44 INFO:       ticket_flights - archive_by_ids 3 rows by ticket_no, flight_id
2020-06-20 19:27:44 INFO:       END ARCHIVE REFERRING TABLES
2020-06-20 19:27:44 INFO: flights - archive_by_ids 3 rows by id
2020-06-20 19:27:44 INFO: flights - END

Finding Dependencies for a Specified Table

A function to find the dependencies of the specified table table. Required parameters are config_pathand table.

After launch, a dictionary will be displayed on the screen, where:

in_refs- dictionary of referring tables to the given one, where key is the name of the table, value is the list of Foreign Key objects ( pk_main- primary key in the main table, pk_ref- primary key in the referring table, fk_ref- name of the column that is the foreign key to the source table);
out_refs - the dictionary of tables to which the given one refers.

$ pggraph get_table_references --config_path config.hw.local.ini --table flights
{'in_refs': {'ticket_flights': [ForeignKey(pk_main='flight_id', pk_ref='ticket_no, flight_id', fk_ref='flight_id')]},
 'out_refs': {'aircrafts': [ForeignKey(pk_main='aircraft_code', pk_ref='flight_id', fk_ref='aircraft_code')],
              'airports': [ForeignKey(pk_main='airport_code', pk_ref='flight_id', fk_ref='arrival_airport'),
                           ForeignKey(pk_main='airport_code', pk_ref='flight_id', fk_ref='departure_airport')]}}

Finding links to strings with specified Primary Key

A function for finding rows in other tables that refer to idstable rows through a Foreign Key table. Required parameters are config_path, tableand ids.

After launch, a dictionary with the following structure will be displayed on the screen:

{
	pk_id_1: {
		reffering_table_name_1: {
			foreign_key_1: [
				{row_pk_1: value, row_pk_2: value},
				...
			], 
			...
		},
		...
	},
	pk_id_2: {...},
	...
}

Call example:

$ pggraph get_rows_references --config_path config.hw.local.ini --table flights --ids 1,2,3
{1: {'ticket_flights': {'flight_id': [{'flight_id': 1,
                                       'ticket_no': '0005432816945'},
                                      {'flight_id': 1,
                                       'ticket_no': '0005432816941'}]}},
 2: {'ticket_flights': {'flight_id': [{'flight_id': 2,
                                       'ticket_no': '0005433101832'},
                                      {'flight_id': 2,
                                       'ticket_no': '0005433101864'},
                                      {'flight_id': 2,
                                       'ticket_no': '0005432919715'}]}},
 3: {'ticket_flights': {'flight_id': [{'flight_id': 3,
                                       'ticket_no': '0005432817560'},
                                      {'flight_id': 3,
                                       'ticket_no': '0005432817568'},
                                      {'flight_id': 3,
                                       'ticket_no': '0005432817559'}]}}}

Usage in code

In addition to running in the console, the library can be used in Python code. Below are examples of calls in the iPython interactive environment.

Archiving a table

>>> from pg_graph.main import setup_logging
>>> setup_logging(log_level='DEBUG')
>>> from pg_graph.api import PgGraphApi
>>> api = PgGraphApi('config.hw.local.ini')
>>> api.archive_table('flights', [4,5])
2020-06-20 23:12:08 INFO: flights - START
2020-06-20 23:12:08 INFO: flights - start archive_recursive 2 rows (depth=0)
2020-06-20 23:12:08 INFO: 	START ARCHIVE REFERRING TABLES
2020-06-20 23:12:08 DEBUG: 	ticket_flights - ForeignKey(pk_main='flight_id', pk_ref='flight_id, ticket_no', fk_ref='flight_id')
2020-06-20 23:12:08 DEBUG: 	SQL('SELECT flight_id, ticket_no FROM bookings.ticket_flights WHERE (flight_id) IN (%s, %s)')
2020-06-20 23:12:08 INFO: 	ticket_flights - start archive_recursive 30 rows (depth=1)
2020-06-20 23:12:08 INFO: 		START ARCHIVE REFERRING TABLES
2020-06-20 23:12:08 DEBUG: 		boarding_passes - ForeignKey(pk_main='flight_id, ticket_no', pk_ref='flight_id, ticket_no', fk_ref='flight_id, ticket_no')
2020-06-20 23:12:08 INFO: 		boarding_passes - archive_by_fk 30 rows by ForeignKey(pk_main='flight_id, ticket_no', pk_ref='flight_id, ticket_no', fk_ref='flight_id, ticket_no')
2020-06-20 23:12:08 DEBUG: 		SQL('CREATE TABLE IF NOT EXISTS bookings.boarding_passes_archive (LIKE bookings.boarding_passes)')
2020-06-20 23:12:08 DEBUG: 		DELETE FROM boarding_passes by FK flight_id, ticket_no - 30 rows
2020-06-20 23:12:08 INFO: 		END ARCHIVE REFERRING TABLES
2020-06-20 23:12:08 INFO: 	ticket_flights - archive_by_ids 30 rows by flight_id, ticket_no
2020-06-20 23:12:08 DEBUG: 	SQL('CREATE TABLE IF NOT EXISTS bookings.ticket_flights_archive (LIKE bookings.ticket_flights)')
2020-06-20 23:12:08 DEBUG: 	DELETE FROM ticket_flights by flight_id, ticket_no - 30 rows
2020-06-20 23:12:08 DEBUG: 	INSERT INTO ticket_flights_archive - 30 rows
2020-06-20 23:12:08 INFO: 	ticket_flights - start archive_recursive 30 rows (depth=1)
2020-06-20 23:12:08 INFO: 		START ARCHIVE REFERRING TABLES
2020-06-20 23:12:08 DEBUG: 		boarding_passes - ForeignKey(pk_main='flight_id, ticket_no', pk_ref='flight_id, ticket_no', fk_ref='flight_id, ticket_no')
2020-06-20 23:12:08 INFO: 		boarding_passes - archive_by_fk 30 rows by ForeignKey(pk_main='flight_id, ticket_no', pk_ref='flight_id, ticket_no', fk_ref='flight_id, ticket_no')
2020-06-20 23:12:08 DEBUG: 		SQL('CREATE TABLE IF NOT EXISTS bookings.boarding_passes_archive (LIKE bookings.boarding_passes)')
2020-06-20 23:12:08 DEBUG: 		DELETE FROM boarding_passes by FK flight_id, ticket_no - 30 rows
2020-06-20 23:12:08 INFO: 		END ARCHIVE REFERRING TABLES
2020-06-20 23:12:08 INFO: 	ticket_flights - archive_by_ids 30 rows by flight_id, ticket_no
2020-06-20 23:12:08 DEBUG: 	SQL('CREATE TABLE IF NOT EXISTS bookings.ticket_flights_archive (LIKE bookings.ticket_flights)')
2020-06-20 23:12:08 DEBUG: 	DELETE FROM ticket_flights by flight_id, ticket_no - 30 rows
2020-06-20 23:12:08 DEBUG: 	INSERT INTO ticket_flights_archive - 30 rows
2020-06-20 23:12:08 INFO: 	ticket_flights - start archive_recursive 30 rows (depth=1)
2020-06-20 23:12:08 INFO: 		START ARCHIVE REFERRING TABLES
2020-06-20 23:12:08 DEBUG: 		boarding_passes - ForeignKey(pk_main='flight_id, ticket_no', pk_ref='flight_id, ticket_no', fk_ref='flight_id, ticket_no')
2020-06-20 23:12:08 INFO: 		boarding_passes - archive_by_fk 30 rows by ForeignKey(pk_main='flight_id, ticket_no', pk_ref='flight_id, ticket_no', fk_ref='flight_id, ticket_no')
2020-06-20 23:12:08 DEBUG: 		SQL('CREATE TABLE IF NOT EXISTS bookings.boarding_passes_archive (LIKE bookings.boarding_passes)')
2020-06-20 23:12:08 DEBUG: 		DELETE FROM boarding_passes by FK flight_id, ticket_no - 30 rows
2020-06-20 23:12:08 INFO: 		END ARCHIVE REFERRING TABLES
2020-06-20 23:12:08 INFO: 	ticket_flights - archive_by_ids 30 rows by flight_id, ticket_no
2020-06-20 23:12:08 DEBUG: 	SQL('CREATE TABLE IF NOT EXISTS bookings.ticket_flights_archive (LIKE bookings.ticket_flights)')
2020-06-20 23:12:08 DEBUG: 	DELETE FROM ticket_flights by flight_id, ticket_no - 30 rows
2020-06-20 23:12:08 DEBUG: 	INSERT INTO ticket_flights_archive - 30 rows
2020-06-20 23:12:08 INFO: 	ticket_flights - start archive_recursive 3 rows (depth=1)
2020-06-20 23:12:08 INFO: 		START ARCHIVE REFERRING TABLES
2020-06-20 23:12:08 DEBUG: 		boarding_passes - ForeignKey(pk_main='flight_id, ticket_no', pk_ref='flight_id, ticket_no', fk_ref='flight_id, ticket_no')
2020-06-20 23:12:08 INFO: 		boarding_passes - archive_by_fk 3 rows by ForeignKey(pk_main='flight_id, ticket_no', pk_ref='flight_id, ticket_no', fk_ref='flight_id, ticket_no')
2020-06-20 23:12:08 DEBUG: 		SQL('CREATE TABLE IF NOT EXISTS bookings.boarding_passes_archive (LIKE bookings.boarding_passes)')
2020-06-20 23:12:08 DEBUG: 		DELETE FROM boarding_passes by FK flight_id, ticket_no - 3 rows
2020-06-20 23:12:08 INFO: 		END ARCHIVE REFERRING TABLES
2020-06-20 23:12:08 INFO: 	ticket_flights - archive_by_ids 3 rows by flight_id, ticket_no
2020-06-20 23:12:08 DEBUG: 	SQL('CREATE TABLE IF NOT EXISTS bookings.ticket_flights_archive (LIKE bookings.ticket_flights)')
2020-06-20 23:12:08 DEBUG: 	DELETE FROM ticket_flights by flight_id, ticket_no - 3 rows
2020-06-20 23:12:08 DEBUG: 	INSERT INTO ticket_flights_archive - 3 rows
2020-06-20 23:12:08 INFO: 	END ARCHIVE REFERRING TABLES
2020-06-20 23:12:08 INFO: flights - archive_by_ids 2 rows by flight_id
2020-06-20 23:12:09 DEBUG: SQL('CREATE TABLE IF NOT EXISTS bookings.flights_archive (LIKE bookings.flights)')
2020-06-20 23:12:09 DEBUG: DELETE FROM flights by flight_id - 2 rows
2020-06-20 23:12:09 DEBUG: INSERT INTO flights_archive - 2 rows
2020-06-20 23:12:09 INFO: flights - END

Finding Dependencies for a Specified Table

>>> from pg_graph.api import PgGraphApi
>>> from pprint import pprint
>>> api = PgGraphApi('config.hw.local.ini')
>>> res = api.get_table_references('flights')
>>> pprint(res)
{'in_refs': {'ticket_flights': [ForeignKey(pk_main='flight_id', pk_ref='flight_id, ticket_no', fk_ref='flight_id')]},
 'out_refs': {'aircrafts': [ForeignKey(pk_main='aircraft_code', pk_ref='flight_id', fk_ref='aircraft_code')],
              'airports': [ForeignKey(pk_main='airport_code', pk_ref='flight_id', fk_ref='arrival_airport'),
                           ForeignKey(pk_main='airport_code', pk_ref='flight_id', fk_ref='departure_airport')]}}

Finding links to strings with specified Primary Key

>>> from pg_graph.api import PgGraphApi
>>> from pprint import pprint
>>> api = PgGraphApi('config.hw.local.ini')
>>> rows = api.get_rows_references('flights', [1,2,3])
>>> pprint(rows)
{1: {'ticket_flights': {'flight_id': [{'flight_id': 1,
                                       'ticket_no': '0005432816945'},
                                      {'flight_id': 1,
                                       'ticket_no': '0005432816941'}]}},
 2: {'ticket_flights': {'flight_id': [{'flight_id': 2,
                                       'ticket_no': '0005433101832'},
                                      {'flight_id': 2,
                                       'ticket_no': '0005433101864'},
                                      {'flight_id': 2,
                                       'ticket_no': '0005432919715'}]}},
 3: {'ticket_flights': {'flight_id': [{'flight_id': 3,
                                       'ticket_no': '0005432817560'},
                                      {'flight_id': 3,
                                       'ticket_no': '0005432817568'},
                                      {'flight_id': 3,
                                       'ticket_no': '0005432817559'}]}}}

The source code of the library is available on GitHub under the MIT license, as well as in the PyPI repository .

Comments, commits and suggestions are welcome.

I will try to answer the questions as far as possible here and in the repository.

PgGraph - utility for archiving and finding table dependencies in PostgreSQL

Background

How to use

Installation and configuration

Running from the console

Parameters

Command examples

Archiving a table

Finding Dependencies for a Specified Table

Finding links to strings with specified Primary Key

Usage in code

Archiving a table

Finding Dependencies for a Specified Table

Finding links to strings with specified Primary Key

More articles: