Your first serverless Python function with OpenFaaS

by Alex Ellis

Let's write a Python function with OpenFaaS that executes hello-world and then move onto something more. We'll start by deploying OpenFaaS on our machine and then we'll install the CLI and write the handler for our function.

OpenFaaS is democratising serverless functions - through Docker containers.

Pre-requisites:

You can read my Introduction to Functions as a Service here.

Step 1 - Install the OpenFaaS CLI

The CLI is a convenient way to interact with your functions. You can use it to build and/or deploy functions to your cluster.

On Linux type in:

$ curl -sSL https://cli.openfaas.com | sudo sh

If you're using Windows, then you're not out of luck, you can actually find the Windows executable on the FaaS releases page.

Step 3 - Deploy OpenFaaS

Deploy OpenFaaS with Kubernetes or with faasd, pick one and follow the instructions: OpenFaaS Deployment

Use the instructions given to you to log in with faas-cli login

You can try out one of the sample functions from the Function Store.

For instance The function named hubstats will tell you how many Docker Hub images a user has. I typed in my user account and found out I have almost 200 already!

To deploy it yourself run:

faas-cli store list

faas-cli store deploy hubstats

Note: this function is not available for Raspberry Pi at this time.

Step 3 - Write your function

Create a new folder for your work:

$ mkdir -p ~/functions && \
  cd ~/functions

Now let's scaffold a new Python function using the CLI:

$ faas-cli new --lang python hello-python

This creates three files for you:

hello-python/handler.py
hello-python/requirements.txt
hello-python.yml

Let's edit the handler.py file:

def handle(req):
    print("Hello! You said: " + req)

handler.py

All your functions should be specified in a YAML file like this - it tells the CLI what to build and deploy onto your OpenFaaS cluster.

Checkout the YAML file hello-python.yml:

provider:
  name: faas
  gateway: http://localhost:8080

functions:
  hello-python:
    lang: python
    handler: ./hello-python
    image: hello-python

hello-python.yml

  • gateway- here we can specify a remote gateway if we need to, what the programming language is and where our handler is located within the filesystem.

  • functions - this block defines the functions in our stack

  • lang: python - even though Docker is used behind the scenes to package your function. You don't have to write your own Dockerfile unless you want to.

  • handler - this is the folder / path to your handler.py file and any other source code you need

  • image - this is the Docker image name. If you are going to push to the Docker Hub change the prefix from hello-python to include your Docker Hub account - i.e. alexellis/hello-python

So let's build the function.

$ faas-cli build -f ./hello-python.yml
...

Successfully tagged hello-python:latest
Image: hello-python built.

You'll now see output from the Docker Engine as it builds your function into an image in your local Docker library. You'll see the image appear on docker images

For example you could run:

$ docker images | grep hello-python
hello-python        latest       e0344b26305f     one minute ago

Testing on a single host

If you're trying thing out on a single host, then you don't need to push your images to a registry, they'll just be used from the local Docker library.

Remote host, or multi-node cluster

If you are using a remote server or a multi-node cluster then you can push your function's image to a registry or the Docker Hub. The Docker Hub is a free service provided by Docker Inc for sharing container images over the Internet.

If you are an advanced user, you can also run a container image registry within your own network. Link: Docker Registry docs

You'll also need to image the image: name to include your Hub account such as image: alexellis2/hello-python.

Here's how to upload the function to a remote registry (if needed):

$ faas-cli push -f ./hello-python.yml

Once you have multiple functions you can also use the --parallel argument to speed things up.

Let's deploy the function:

$ faas-cli deploy -f ./hello-python.yml
Deploying: hello-python.
No existing service to remove
Deployed.
200 OK
URL: http://localhost:8080/function/hello-python

And it's ready to be tested! Either open up the UI or use curl to call it:

$ curl localhost:8080/function/hello-python -d "it's Alex here"
Hello! You said: its Alex here

You can even use the faas-cli to list and invoke your functions.

Try playing with these two commands including looking at the help page with the --help parameter:

  • faas-cli list
  • echo "Hello!" | faas-cli invoke hello-python

Step 4 - Import 3rd party dependencies

So what if you need some dependencies to add behaviour beyond the standard Python libraries? Well you can use pip by providing a requirements.txt file along with your function handler.

Let's include the popular requests module which we can use to interact with webpages.

You will see a file called hello-python/requirements.txt. Add the following line:

requests

requirements.txt

Now we can update our Python code. Let's make it so it can accept JSON request of a URL and a string we want to test for:

We'll trigger it using a JSON request, which will take this format:

{
 "url": "https://blog.alexellis.io/rss/",
 "term": "docker"
}

Now update the hello-python/handler.py file:

import requests
import json

def handle(req):
    result = {"found": False}
    json_req = json.loads(req)
    r = requests.get(json_req["url"])
    if json_req["term"] in r.text:
        result = {"found": True}

    print json.dumps(result)

handler.py

Then just rebuild, (re-push if needed) and re-deploy.

$ faas-cli build -f ./hello-python.yml && \
  faas-cli deploy -f ./hello-python.yml

*This time you'll see pip installing your function's modules such as requests.

Finally test it out!

$ curl localhost:8080/function/hello-python --data-binary '{
 "url": "https://blog.alexellis.io/rss/",
 "term": "docker"
}'

Here's the result:

{"found": true}

Let's try another site:

$ curl localhost:8080/function/hello-python --data-binary '{
 "url": "https://www.kubernetes.io/",
 "term": "docker"
}'
{"found": False}

You can also access the function just as easily via the UI.

  • Metrics

Prometheus is also baked into the FaaS stack, which means you can checkout the various metrics on how your functions are being used as well as how long they're taking to run. You can view the Prometheus UI at http://localhost:9090

  • Troubleshooting

If you run into any errors such as Can't reach service check the following:

  • If it's a multi-node cluster, you have to push your images to the Docker Hub or similar registry before deploying
  • If you're still getting the error docker service logs --no-trunc=true hello-python will give a bit more info.

Step 5 - Going further

Please show support and Star the project on Github

Now that you've built your first function, why not checkout some of the examples built by the OpenFaaS community? You can share your first serverless OpenFaaS function with us on Twitter or Github.

Want to know more about the OpenFaaS Serverless framework for containers?

You can read my Introduction to Functions as a Service here.

Setup FaaS with free credits

Learn Docker

If you want to learn more about Docker for DevOps and developers, checkout my Docker series.