How to scale the app in Kubernetes with KEDA

Installing KEDA

you can use any Kubernetes. Installing KEDA is simple, I use helm:

helm repo add kedacore https://kedacore.github.io/charts 
helm repo update
helm install keda kedacore/keda --namespace keda --create-namespace

Now check that the KEDA and metrics server are up:

kubectl get po -n keda

kubectl get apiservices v1beta1.external.metrics.k8s.io

That’s it, KEDA operator is installed, let’s move on.

Let’s deploy “Hello World”

kubectl create ns nginx-demo

kubectl apply -n nginx-demo -f https://raw.githubusercontent.com/kubernetes/website/main/content/en/examples/controllers/nginx-deployment.yaml

Everything is good, this deployment has 3 pods, as it should.

Let’s scale the pods depending on the number of messages in the RabbitMQ queue.

Let’s simulate a real situation where RabbitMQ is not running inside Kubernetes but in a different network.  But before that, let’s find the IP address of local machine, we will need it soon:

ipconfig getifaddr eth0 
192.168.1.101

All right, let’s install RabbitMQ and start it:

apt install rabbitmq 
RABBITMQ_NODE_IP_ADDRESS=0.0.0.0 /usr/local/sbin/rabbitmq-server

Let’s make sure everything is up:

/usr/local/sbin/rabbitmqadmin -H 127.0.0.1 -u guest -p guest \ list connections 
No items

Everything is good. The only thing is that access for the default guest user is only available at 127.0.0.1, so we need to add a new user (demo/demo) and create a queue (demo_queue) for testing:

/usr/local/sbin/rabbitmqadmin --host 127.0.0.1 -u guest -p guest \ declare user name=demo password=demo tags=administrator user declared 

/usr/local/sbin/rabbitmqadmin --host 127.0.0.1 -u guest -p guest \ declare permission vhost=/ user=demo configure=".*" write=".*" read=".*" permission declared 

/usr/local/sbin/rabbitmqadmin --host 127.0.0.1 -u guest -p guest \ declare queue name=demo_queue queue declared

After that, we can connect via the “external” IP address:

/usr/local/sbin/rabbitmqadmin -H 192.168.1.101 -u demo -p demo \ list queues

RabbitMQ is ready, now we need to connect KEDA to it. For this, we need to deploy three objects:

Secret – it will store the connection string to RabbitMQ: amqp://demo:demo@192.168.1.101:5672/

TriggerAuthentication – an authentication object that uses the data from the secret above

ScaledObject – the object where we can configure various scaling parameters

Deploy:

kubectl apply -n nginx-demo -f objects.yaml

---
apiVersion: v1
kind: Secret
metadata:
  name: keda-rabbitmq-secret
data:
  host: YW1xcDovL2RlbW86ZGVtb0AxOTIuMTY4LjIuMTAxOjU2NzIv # echo -n amqp://demo:demo@192.168.1.101:5672/ | base64
---
apiVersion: keda.sh/v1alpha1
kind: TriggerAuthentication
metadata:
  name: keda-trigger-auth-rabbitmq-conn
spec:
  secretTargetRef:
    - parameter: host
      name: keda-rabbitmq-secret 
      key: host
---
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: rabbitmq-scaledobject
spec:
  scaleTargetRef:
    name: nginx-deployment 
  minReplicaCount: 1       
  maxReplicaCount: 5       
  pollingInterval: 10      
  cooldownPeriod:  120     
  triggers:
  - type: rabbitmq
    metadata:
      protocol: amqp
      queueName: demo_queue 
      mode: QueueLength
      value: "3"
    authenticationRef:
      name: keda-trigger-auth-rabbitmq-conn

Let’s check that everything is working:

kubectl get scaledobject 
kubectl get po 
kubectl get hpa

Testing

Let’s send 5 messages to our RabbitMQ queue:

for i in {1..5}; do
    /usr/local/sbin/rabbitmqadmin --host 192.168.1.101 -u demo -p demo \
        publish exchange=amq.default routing_key=demo_queue payload="message ${i}"
done

/usr/local/sbin/rabbitmqadmin -H 192.168.1.101 -u demo -p demo \
    list queues

Wait couple of seconds while KEDA fetches metrics and checks our pods:

kubectl get hpa

kubectl get po

Let’s add more messages to the queue:

for i in {6..12}; do
    /usr/local/sbin/rabbitmqadmin --host 192.168.1.101 -u demo -p demo \
        publish exchange=amq.default routing_key=demo_queue payload="message ${i}"
done

/usr/local/sbin/rabbitmqadmin -H 192.168.1.101 -u demo -p demo \
    list queues

12 messages in the queue, theoretically, 4 pods should be up, let’s check:

kubectl get hpa

kubectl get po

Now let’s clear the queue, the number of pods should drop to the minimum value:

/usr/local/sbin/rabbitmqadmin -H 192.168.1.101 -u demo -p demo \
    purge queue name=demo_queue
queue purged

/usr/local/sbin/rabbitmqadmin -H 192.168.1.101 -u demo -p demo \
    list queues

Everything is good, it scales up and down as expected.

Let’s clean up:

helm uninstall keda -n keda
kubectl delete ns keda
kubectl delete ns nginx-demo