Azure Functions in Kubernetes Example

2022-03-31 700 words 4 minutes

Contents

Introduction

This tutorial shows how to deploy two simple examples of Azure Functions written in Python to Kubernetes (AKS used here, but not required).

Envrironment setup

Install prerequisites

Docker (e.g. https://www.docker.com/products/docker-desktop)
Kubectl (https://kubernetes.io/docs/tasks/tools/install-kubectl/)
Azure cli (https://docs.microsoft.com/cs-cz/cli/azure/install-azure-cli)
Func cli (https://docs.microsoft.com/en-us/azure/azure-functions/functions-run-local)

Download Functions source code

git clone git@github.com:pajikos/cloud-samples.git
cd cloud-samples/azure-functions-in-k8s

Setup your custom properties

Edit file .env and fill correct values.

Load envrironment properties

# Load env properties
source .env

You need a Docker registry with push rights to be able to upload images with two functions. You have to be logged, here is an example when using Azure Container Registry:

# Auth with AAD accounts
az acr login -n $ACR_NAME

Deploy KEDA to cluster

For more info, visit: https://keda.sh/docs/2.6/deploy/ Please note, that you do not need Keda version 2.2 and above(required for Durable Functions only).

Create namespace for Keda

# Create namespace
kubectl create namespace keda

There are two ways how to deploy KEDA:

Install Keda using Azure Func Core Tools

# Deploy keda
func kubernetes install --namespace keda

Install Keda using the official Helm chart

# Add Helm repo
helm repo add kedacore https://kedacore.github.io/charts
#Update Helm repo
helm repo update
# Create namespace
kubectl create namespace keda
#Install keda Helm chart
helm install keda kedacore/keda --namespace keda

Deploying Azure Functions (non-durable function)

This example shows how to deploy two functions to the AKS cluster:

Simple HTTP Trigger-based function - return response to HTTP GET imediately.
ServiceBus trigger-based function - automatic scaling based on the number of messages in a queue, simply consuming messages and logging them to console

Setup ServiceBus Namespace

One of the functions uses Azure Service Bus, so here is the tutorial on how to create it using az cli:

az servicebus namespace create --resource-group $RG --name $SERVICEBUS_NAMESPACE --location $LOC --sku Standard

az servicebus queue create --name $QUEUE_NAME \
    --resource-group $RG \
    --namespace-name $SERVICEBUS_NAMESPACE

az servicebus queue authorization-rule create --resource-group $RG --namespace-name $SERVICEBUS_NAMESPACE --queue-name $QUEUE_NAME --name $POLICY_NAME --rights Listen Send Manage
PRIMARY_KEY=$(az servicebus queue authorization-rule keys list --resource-group $RG --namespace-name $SERVICEBUS_NAMESPACE --queue-name queue-input --name $POLICY_NAME --query primaryKey --output tsv)   
# Assign env property with connection strings to queue
export ServiceBusConnectionString="Endpoint=sb://$SERVICEBUS_NAMESPACE.servicebus.windows.net/;SharedAccessKeyName=$POLICY_NAME;SharedAccessKey=$PRIMARY_KEY;"
# Keda needs Connection string including entity path (i.e. the full path)
export ServiceBusConnectionStringQueue="Endpoint=sb://$SERVICEBUS_NAMESPACE.servicebus.windows.net/;SharedAccessKeyName=$POLICY_NAME;SharedAccessKey=$PRIMARY_KEY;EntityPath=$QUEUE_NAME"
# Run the following lines and insert the output to local.settings.json 
echo "\"ServiceBusConnectionStringQueue\": \"$ServiceBusConnectionStringQueue\""
echo "\"ServiceBusConnectionString\": \"$ServiceBusConnectionString\""


# Optional, Getting root access key (= global access to the whole namespace)
ROOT_CONNECTION_STRING=$(az servicebus namespace authorization-rule keys list --resource-group $RG --namespace-name $SERVICEBUS_NAMESPACE --name RootManageSharedAccessKey --query primaryConnectionString --output tsv)

Test local run

If the Azure Service Bus and queue successfully installed, you can run functions locally:

cd simple_function_demo
func start

If no problem, successful output:

[azure-functions-in-k8s/simple_function_demo]func start
Found Python version 3.9.10 (python3).

Azure Functions Core Tools
Core Tools Version:       4.0.3971 Commit hash: d0775d487c93ebd49e9c1166d5c3c01f3c76eaaf  (64-bit)
Function Runtime Version: 4.0.1.16815

...

Functions:

  HttpTriggerTest: [GET,POST] http://localhost:7071/api/HttpTriggerTest

  ServiceBusQueueTrigger: serviceBusTrigger

Building and deploying the testing function

If run local properly, we can now deploy functions to AKS.

FUNCTION_NAME=k8sfntest
FUNCTION_NS_NAME=functions-test
func kubernetes deploy --name $FUNCTION_NAME --min-replicas 0 --max-replicas 5 --cooldown-period 30 --image-name $ACR_NAME.azurecr.io/functiontestdeploy:latest --namespace $FUNCTION_NS_NAME -i --dry-run > deployment.yaml

The following steps are needed:

Open deployment.yaml for editing
Find definition of ScaledObject (k8sfntest) and replace: ServiceBusConnectionString -> ServiceBusConnectionStringQueueInput (Kuda Azure ServiceBus integration needs full queue endpoint.)

# Create namespace
kubectl create namespace $FUNCTION_NS_NAME
# Run deployment
kubectl apply -f deployment.yaml -n $FUNCTION_NS_NAME

Setting access to deployed app

A new service during deployment was created, so here you can extract its external service IP.

# list all running services
kubectl get services --namespace $FUNCTION_NS_NAME
# Save service IP to variable
SERVICE_IP=$(k get svc $FUNCTION_NAME-http --namespace $FUNCTION_NS_NAME -o jsonpath='{.status.loadBalancer.ingress[*].ip}')

Go to:

echo "Invoke url: http://$SERVICE_IP/api/httptriggertest?name=Pavel"

Test consuming messages from the queue

To send a batch of messages to the Azure Service Bus namespace, you can do it on Azure Portal or simply use my Python script. Install script dependencies first:

pip install azure.servicebus

In the Azure Service Bus section of this page, a connection string was exported as an environment variable, so you can now run the script directly:

python send_message.py

Successful output:

Sent a list of 100 messages
Sent a batch of 100 messages
Done sending messages
-----------------------

Clean resources

kubectl delete namespace $FUNCTION_NS_NAME