ktweeter 🔊

An example of an Azure Function that can post K8s events to a Twitter feed running in Kubernetes and using Knative.

Setup

Prerequisites:

• Download and install Azure Functions CLI func. https://github.com/Azure/azure-functions-core-tools#installing
• Python 3.6
• Setup OpenShift/Kubernetes as described below with Knative.

On OpenShift

Make sure you have a cluster Knative installed or follow these instructions.

On Minikube

Follow the instructions here to setup Minikube and Knative.

Egress traffic (Internet connectivity)

In order to allow egress traffic you need to configure one of the Knative Serving config maps. An example of how to do configure minikube is avaible on the patches folder and can be applied as follows:

kubectl patch configmap/config-network -n knative-serving --type json --patch "$(cat patches/patch_config-network.json)"

This is required for the function to post on Twitter. For more information please check the examples here.

Create a function project

For detailed information about how to create Python Functions in Azure Functions follow this guide: https://docs.microsoft.com/en-us/azure/azure-functions/functions-create-first-function-python

• Create a new python function app

  func init --worker-runtime python --docker

• Create a http-triggered function

  func new --template 'HTTP trigger' --name http-trigger

• Enable function logging to console

  cat <<EOF > host.json
  {
      "version":  "2.0",
      "logging": {
          "console": {
              "isEnabled": "true"
          },
          "logLevel": {
              "default": "Information"
          }
      },
  }
  EOF
  

• Disable auth for our http trigger

  sed -i -e 's/"authLevel": "function"/"authLevel": "anonymous"/' http-trigger/function.json

The sample code to send cloudevents to a Twitter feed is on http-trigger/__init__.py

Deploy

• Set container registry

  export REGISTRY=docker.io/markito

• Update the Dockerfile - Add the following line right after FROM

  ENV ASPNETCORE_URLS=http://+:8080

• Deploy the function to Knative

  func deploy --platform knative --name http-trigger --registry $REGISTRY --config ~/.kube/config

• Deploy the secret file with your Twitter API token and keys

For example:

apiVersion: v1
kind: Secret
metadata:
  name: twitter.creds
type: Opaque
data:
  consumer_key: YWNjZXNzX3Rva2VuX3NlY3JldA==
  consumer_secret: YWNjZXNzX3Rva2VuX3NlY3JldA==
  access_token: YWNjZXNzX3Rva2VuX3NlY3JldA==
  access_token_secret: YWNjZXNzX3Rva2VuX3NlY3JldA==

There is a twitter.yaml file that should be updated with proper values. Once you update the file you can create the secrets using:

kubectl apply -n azure-functions -f twitter.yaml

Remember to encode those values in base64. For example: echo -n 'VALUE' | base64 VkFMVUU=

• We need to add the secrets to the service specification. The easiest way to do that is to apply a patch as follows:

  kubectl patch ksvc/http-trigger -n azure-functions --type json --patch "$(cat patches/env_patch.json)"
  service.serving.knative.dev/http-trigger patched
  

This patch will generate a new Revision in the Knative service since we are modifying the configuration of the application.

• Ensure the function is up and returns a 200

  curl -v -H "Host: httptrigger.azure-functions.example.com" http://$(minikube ip):31380/api/http-trigger

• Simulating POSTing an empty CloudEvent to it

  curl -d "{}" -H "Content-Type: application/json" -H "Host: httptrigger.azure-functions.example.com" http://$(minikube ip):31380/api/http-trigger

• Check out the function logs

  kubectl logs -n azure-functions -l serving.knative.dev/service=httptrigger -c user-container

Knative Event Source Setup

Create a ServiceAccount with permissions to watch k8s events

kubectl apply -f eventing/serviceAccount.yaml

Create a Knative Eventing Channel

kubectl apply -f eventing/channel.yaml

Create a KubernetesEventSource

kubectl apply -f eventing/k8sEventSource.yaml

Create a Subscription pointing to our Azure Function

kubectl apply -f eventing/subscription.yaml

Sending events

• Create a Kubernetes event to confirm things work kubectl run -n default -i --tty busybox --image=busybox --restart=Never --rm=true -- sh

Just exit the container after it starts by typing exit

• Check the Twitter feed or container logs using:

  kubectl logs -n azure-functions -l serving.knative.dev/service=httptrigger -c user-container

You should see something like:

info: Function.http-trigger[0]
      Executed 'Functions.http-trigger' (Succeeded, Id=4f07df31-a12c-4acd-9ede-297887fb38f7)
info: Function.http-trigger[0]
      Executing 'Functions.http-trigger' (Reason='This function was programmatically called via the host APIs.', Id=4b20130f-0869-4a16-aff1-11da9a447dbd)
info: Function.http-trigger.User[0]
      {'metadata': {'name': 'busybox.158e659be89c8e31', 'namespace': 'default', 'selfLink': '/api/v1/namespaces/default/events/busybox.158e659be89c8e31', 'uid': 'acdf5462-4cea-11e9-93fd-307a181b66b4', 'resourceVersion': '31059', 'creationTimestamp': '2019-03-22T21:37:21Z'}, 'involvedObject': {'kind': 'Pod', 'namespace': 'default', 'name': 'busybox', 'uid': '955d19f5-4cea-11e9-93fd-307a181b66b4', 'apiVersion': 'v1', 'resourceVersion': '30978', 'fieldPath': 'spec.containers{busybox}'}, 'reason': 'Killing', 'message': 'Killing container with id docker://busybox:Need to kill Pod', 'source': {'component': 'kubelet', 'host': 'minikube'}, 'firstTimestamp': '2019-03-22T21:37:21Z', 'lastTimestamp': '2019-03-22T21:37:21Z', 'count': 1, 'type': 'Normal', 'eventTime': None, 'reportingComponent': '', 'reportingInstance': ''}
info: Function.http-trigger[0]
      Executed 'Functions.http-trigger' (Succeeded, Id=4b20130f-0869-4a16-aff1-11da9a447dbd)

Troubleshooting

• Status is a duplicate.

When monitoring the logs if you see the following message:

Error processing event or posting update.
      Message: [{'code': 187, 'message': 'Status is a duplicate.'}]

That's because Twitter's API don't allow for duplicate messages within a certain period of time.

• Old Python version being detected.

Python 3.6 is required. Current python version is 'Python 2.7.15

Workaround: Create a Python environment using the new version. For example:

source .env/bin/activate