SQLAlchemy through Paramiko SSH

In case there is anyone who's interested in connecting to a remote Postgresql database via SSH and wants to load data into a pandas DataFrame here is how to do it.

Suppose we have installed a postgresql database on a remote server, to which we can ssh by the following parameters.

SSH parameters:

• Server's ip: 10.0.0.101
• SSH port: 22 (default port for SSH)
• Username: my_username
• Password: my_password

Database parameters:

• Port: 5432 (postgresql default port)
• Database name: db
• Database user: postgres_user (default username is postgres)
• Database password: postgres_pswd (default password is an empty string)
• Table with our data: MY_TABLE

Now, we want to connect to this database on our end and load data into a pandas DataFrame:

from sshtunnel import SSHTunnelForwarder
from sqlalchemy import create_engine
import pandas as pd

server = SSHTunnelForwarder(
    ('10.0.0.101', 22),
    ssh_username="my_username",
    ssh_password="my_password",
    remote_bind_address=('127.0.0.1', 5432)
    )

server.start()
local_port = str(server.local_bind_port)
engine = create_engine('postgresql://{}:{}@{}:{}/{}'.format("postgres_user", "postgres_pswd", "127.0.0.1", local_port, "db"))

dataDF = pd.read_sql("SELECT * FROM \"{}\";".format("MY_TABLE"), engine)

server.stop()

You could use the SSHTunnel library as follows:

from sshtunnel import SSHTunnelForwarder #Run pip install sshtunnel
from sqlalchemy.orm import sessionmaker #Run pip install sqlalchemy

with SSHTunnelForwarder(
    ('10.160.1.24', 22), #Remote server IP and SSH port
    ssh_username = "<usr>",
    ssh_password = "<pwd>",
    remote_bind_address=('127.0.0.1', 5432)
    ) as server:

    server.start() #start ssh sever
    print 'Server connected via SSH'

    #connect to PostgreSQL
    local_port = str(server.local_bind_port)
    engine = create_engine('postgresql://<db_user>:<db_pwd>@127.0.0.1:' + local_port +'/<db_name>')

    Session = sessionmaker(bind=engine)
    session = Session()

    print 'Database session created'

    #test data retrieval
    test = session.execute("SELECT * FROM <table_name>")

The easiest way to do this would be to run an SSH tunnel to the mysql port on the remote host. For example:

ssh -f [email protected] -L 3307:mysql1.example.com:3306 -N

Then connect locally with SQLAlchemy:

engine = create_engine("mysql://username_sql:password_sql@localhost:3307/dbb")

If you really want to use paramiko, try this demo code in the paramiko repo or the sshtunnel module. The ssh command might be the easiest method though.. and you can use autossh to restart the tunnel if it goes down.