Context manager to run PostgreSQL queries with SQLAlchemy, terminating any other clients that continue to block it after a configurable delay.

Using this to wrap queries is somewhat of a last resort, but is useful in certain Extract Transform Load (ETL) pipeline contexts. For example, if it is more important to replace one table with another than to allow running queries on the table to complete, then this can be used to run the relevant ALTER TABLE RENAME TO query.

pg-force-execute can be installed from PyPI using pip. psycopg2 or psycopg (Psycopg 3) must also be explicitly installed.

pip install pg-force-execute psycopg

import datetime
import sqlalchemy as sa
from pg_force_execute import pg_force_execute

# Run postgresql locally should allow the below to run
# docker run --rm -it -e POSTGRES_HOST_AUTH_METHOD=trust -p 5432:5432 postgres

engine = sa.create_engine('postgresql+psycopg://[email protected]:5432/')
query = 'SELECT 1'  # A more realistic example would be something that needs an exclusive lock on a table

with \
        engine.begin() as conn, \
        pg_force_execute(
            conn,                                 # SQLAlchemy connection to run the query
            delay=datetime.timedelta(minutes=5),  # Amount of time to wait before cancelling queries
        ):

    results = conn.execute(sa.text(query))
    print(results.fetchall())

The API a single context manager pg_force_execute.

pg_force_execute(conn, delay=datetime.timedelta(minutes=5), check_interval=datetime.timedelta(seconds=1), cleanup_timeout=datetime.timedelta(seconds=10), logger=logging.getLogger("pg_force_execute"))

• conn - A SQLAlchemy connection that will be unblocked

• delay (optional) - How long to wait before attempting to terminate backends blocking conn

• check_interval (optional) - The interval between repeated attempts to terminate backends blocking conn

• cleanup_timeout (optional) - How long to wait for resources to be cleaned up before allowing exit of the context manager

  For usual operation this parameter shouldn't need to be changed.

• logger (optional) The Python logger instance through which to log

• Python >= 3.7.1 (tested on 3.7.1, 3.8.0, 3.9.0, 3.10.0, and 3.11.0)
• psycopg2 >= 2.9.2 or Psycopg 3 >= 3.1.4
• SQLAlchemy >= 1.4.24 (tested on 1.4.24 and 2.0.0)
• PostgreSQL >= 9.6 (tested on 9.6, 10.0, 11.0, 12.0, 13.0, 14.0, 15.0, and 16 Beta 2)

Note that SQLAlchemy < 2 does not support Psycopg 3.

python -m pip install -e ".[dev]"  # Only needed once
./start-services.sh                # Only needed once
pytest