SQLAlchemy

PostgreSQL

db.py

import os
from sqlalchemy import create_engine

engine = create_engine(
    "postgresql+psycopg://app:%s@pg-7f3k.us-east.databasezy.com:5432/app" % os.environ["ZB_PASSWORD"],
    connect_args={"sslmode": "verify-full", "sslrootcert": "databasezy-ca.pem"},
    pool_size=5,
    max_overflow=0,
    pool_pre_ping=True,     # detects connections dropped by a free-tier sleep
    pool_recycle=1800,
)

Async: create_async_engine("postgresql+psycopg://...") takes the same connect_args.

MySQL

db.py

import os
from sqlalchemy import create_engine

engine = create_engine(
    "mysql+pymysql://app:%s@mysql-7f3k.us-east.databasezy.com:3306/app?charset=utf8mb4" % os.environ["ZB_PASSWORD"],
    connect_args={"ssl": {"ca": "databasezy-ca.pem"}, "ssl_verify_cert": True, "ssl_verify_identity": True},
    pool_size=5,
    max_overflow=0,
    pool_pre_ping=True,
    pool_recycle=1800,
)

Alembic

alembic.ini (excerpt)

sqlalchemy.url = postgresql+psycopg://app:%(ZB_PASSWORD)[email protected]:5432/app?sslmode=verify-full&sslrootcert=databasezy-ca.pem

Run alembic upgrade head from CI or a Kubernetes Job against the direct endpoint. With the pooled endpoint at runtime, set prepare_threshold=None on psycopg (connect_args) so prepared statements are not reused across pooled sessions.