faas/guide/self_hosted_registry.md

Use a self-hosted registry with OpenFaaS

If you're using OpenFaaS on single host, then you don't need to push your images to a registry. They will just be used from the local Docker library.

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.

This describes how to use OpenFaaS in a swarm with your own local registry for hosting function images.

Set up a swarm

For this example lets presume you want to create a swarm of 3 nodes. Use node1 as the manager.

This is adapted from the Swarm deployment guide.

$ docker swarm init --advertise-addr $(hostname -i)

Now in node2 and node3 paste the output from the above command.

$ docker swarm join --token ...

Install OpenFaaS

$ git clone https://github.com/openfaas/faas && \
  cd faas && \
  ./deploy_stack.sh

Start a registry

Add it to the swarm and use the same network as OpenFaaS.

docker service create -d -p 5000:5000 \
  --network func_functions --detach=false \
  --name registry registry:2

Here we are using a basic local registry. You can deploy it elsewhere and use volumes depending on your persistence needs. If you would like to use authentication with your registry, this guide may be helpful.

Install the CLI

This is a helper for using and deploying functions to OpenFaaS.

On a Mac if you're using brew then you can type in

$ brew install faas-cli

On Linux

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

Create a function

$ mkdir -p ~/functions/hello-python
$ cd ~/functions

hello-python/handler.py

import socket
def handle(req):
    print("Hello world from " + socket.gethostname())

stack.yml

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

functions:  
  hello-python:
    lang: python
    handler: ./hello-python/
    image: localhost:5000/faas-hello-python

Let's build the function

$ faas-cli build -f ./stack.yml

Upload the function to our registry

$ faas-cli push -f ./stack.yml

Check that the image made it to the registry

$ curl localhost:5000/v2/faas-hello-python/tags/list
{"name":"faas-hello-python","tags":["latest"]}

Now we will delete the local image to be sure the deployment happens from the registry

$ docker rmi localhost:5000/faas-hello-python
$ docker images | grep hello | wc -l
0

Deploy the function from the registry

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

See that the function works

$ curl -X POST localhost:8080/function/hello-python
Hello world from 281c2858c673

Update the function

hello-python/handler.py:

import socket
def handle(req):
    print("Hello EARTH from " + socket.gethostname())

Now we can rebuild, push and deploy it to the swarm.

$ faas-cli build -f ./stack.yml && \
  faas-cli push -f ./stack.yml && \
  faas-cli deploy -f ./stack.yml

See that the update works

$ curl -X POST localhost:8080/function/hello-python
Hello EARTH from 9dacd2333c1c

Scale it

Start some replicas so there are functions spanning the swarm

$ docker service scale hello-python=3 --detach=false
hello-python scaled to 3
overall progress: 3 out of 3 tasks
1/3: running
2/3: running
3/3: running

$ docker service ls | grep hello-python
oqvof6b9gpvl        hello-python        replicated          3/3                 localhost:5000/faas-hello-python:latest

Test the function across the swarm

$ curl -X POST localhost:8080/function/hello-python
Hello earth from 9dacd2333c1c
$ curl -X POST localhost:8080/function/hello-python
Hello earth from 281c2858c673
$ curl -X POST localhost:8080/function/hello-python
Hello earth from 9dacd2333c1c