Using MongoDB on PythonAnywhere

Getting a MongoDB server

We don't provide Mongo servers ourselves, so you'll need to get one from an external provider (many of our customers are using MongoDB Atlas). For best performance, you should provision your server in the AWS us-east-1 datacenter if you're using our US-based site at www.pythonanywhere.com, or the eu-central-1 datacenter if you're using our EU site at eu.pythonanywhere.com.

Because this will require non-HTTP external Internet access, you'll need a paid account.

There is a complete step-by-step tutorial on our blog.

Connecting to it.

If you're connecting from a console or a scheduled task, just use the regular PyMongo, creating a MongoClient object with the normal parameters to specify the server, the username and the password..

If you're connecting from a web app, and you're using Flask, we recommend that you use the Flask-PyMongo extension, which works well in a multiprocessing environment like websites on our system.

If you're just using the "raw" PyMongo API, there are a few extra parameters you need to to add to your call to pymongo.MongoClient:

For pymongo < 4:

connectTimeoutMS=30000, socketTimeoutMS=None, socketKeepAlive=True, connect=False, maxPoolsize=1

For pymongo >= 4:

connectTimeoutMS=30000, socketTimeoutMS=None, connect=False, maxPoolsize=1

This handles the bulk of the stuff that would otherwise be handled by Flask-PyMongo if you were using it.

Accesslist your IP address for MongoDB Atlas

MongoDB Atlas requires you to provide the IP address of any code that is trying to connect to your Mongo instance; this can be a bit tricky on PythonAnywhere, because the precise IP address will depend on the time your code runs, and whether it's running in a website's code, or in a task, or in a console.

The easiest, solution, though certainly not the most secure, is to accesslist the CIDR 0.0.0.0/0, which is a "accesslist" containing every IP address on the Internet.

A more secure solution is to use the MongoDB Atlas API to tell them about new IP addresses that they should allow to connect. You can combine that with the ipify service, which tells you what IP address your code is using right now, to make your code automatically accesslist the IP it's running on when it starts up. The following code (based on code provided by Nicolas Oteiza and modified by Amitoz Azad) will do that. You just need to install the ipify module and then use this, replacing the bits inside the <>s:

import requests
from requests.auth import HTTPDigestAuth
from ipify import get_ip

atlas_group_id = "<your group ID aka project ID -- check the Project / Settings section inside Atlas>"
atlas_api_key_public = "<your atlas public API key>"
atlas_api_key_private = "<your atlas private API key>"
ip = get_ip()

resp = requests.post(
    "https://cloud.mongodb.com/api/atlas/v1.0/groups/{atlas_group_id}/accessList".format(atlas_group_id=atlas_group_id),
    auth=HTTPDigestAuth(atlas_api_public_key, atlas_api_private_key),
    json=[{'ipAddress': ip, 'comment': 'From PythonAnywhere'}]  # the comment is optional
)
if resp.status_code in (200, 201):
    print("MongoDB Atlas accessList request successful", flush=True)
else:
    print(
        "MongoDB Atlas accessList request problem: status code was {status_code}, content was {content}".format(
            status_code=resp.status_code, content=resp.content
        ),
        flush=True
    )

If there is not a version of the ipify module that works with the version of Python that you are using, you can just use the REST interface to ipify. There is a complete example on the "Code Samples" page on the ipify web site.

If there are any problems accesslisting your IP address (for example, if the API key is wrong) then you will find the error messages in the program's output; for websites, it will be in the server log file (linked from the "Web" page).