Best practices for balancing between Google Cloud SQL postgres read replicas for App Engine App

Question

We are making a move towards cloud deployment right now, and in that process would like to create some read replicas for our database.

I am confused as to what the suggested method for balancing requests between Cloud SQL read replicas. It seems that the Cloud SQL Proxy may be the correct method, but it is unclear to me how it should be used in a production environment.

Answer 1

As you can see here (Does Cloud SQL provide load balancing for requests to read replicas?) "Cloud SQL does not automatically provide or configure load balancing. You can use connection pooling to distribute read requests between read replicas by switching between multiple connection endpoints".

Your code should look like this in case your want to use Unix sockets

# Remember - storing secrets in plaintext is potentially unsafe. Consider using
# something like https://cloud.google.com/secret-manager/docs/overview to help keep
# secrets secret.
db_user = os.environ["DB_USER"]
db_pass = os.environ["DB_PASS"]
db_name = os.environ["DB_NAME"]
db_socket_dir = os.environ.get("DB_SOCKET_DIR", "/cloudsql")
cloud_sql_connection_name = os.environ["CLOUD_SQL_CONNECTION_NAME"]

pool = sqlalchemy.create_engine(

    # Equivalent URL:
    # postgres+pg8000://<db_user>:<db_pass>@/<db_name>
    #                         ?unix_sock=<socket_path>/<cloud_sql_instance_name>/.s.PGSQL.5432
    sqlalchemy.engine.url.URL(
        drivername="postgresql+pg8000",
        username=db_user,  # e.g. "my-database-user"
        password=db_pass,  # e.g. "my-database-password"
        database=db_name,  # e.g. "my-database-name"
        query={
            "unix_sock": "{}/{}/.s.PGSQL.5432".format(
                db_socket_dir,  # e.g. "/cloudsql"
                cloud_sql_connection_name)  # i.e "<PROJECT-NAME>:<INSTANCE-REGION>:<INSTANCE-NAME>"
        }
    ),
    **db_config
)

And the following example is for TCP connections

# Remember - storing secrets in plaintext is potentially unsafe. Consider using
# something like https://cloud.google.com/secret-manager/docs/overview to help keep
# secrets secret.
db_user = os.environ["DB_USER"]
db_pass = os.environ["DB_PASS"]
db_name = os.environ["DB_NAME"]
db_host = os.environ["DB_HOST"]

# Extract host and port from db_host
host_args = db_host.split(":")
db_hostname, db_port = host_args[0], int(host_args[1])

pool = sqlalchemy.create_engine(
    # Equivalent URL:
    # postgres+pg8000://<db_user>:<db_pass>@<db_host>:<db_port>/<db_name>
    sqlalchemy.engine.url.URL(
        drivername="postgresql+pg8000",
        username=db_user,  # e.g. "my-database-user"
        password=db_pass,  # e.g. "my-database-password"
        host=db_hostname,  # e.g. "127.0.0.1"
        port=db_port,  # e.g. 5432
        database=db_name  # e.g. "my-database-name"
    ),
    **db_config
)