keda-demo

• keda-demo
  • Setup
    • Fetch the Helm Repositories
    • Install KEDA
    • Install the Metrics Exporter Workloads (PodInfo)
    • Install a Deployment
  • Demos
    • Autoscaling a Deployment with a Prometheus Metrics
      • Prerequisites
      • Steps
    • Autoscaling a Deployment with a Redis List
      • Steps
    • Autoscaling Jobs with a Redis List
      • Steps

Setup

These are the steps required for setting up and running the demos.

Fetch the Helm Repositories

Get the required helm repositories with the commands below:

helm repo add kedacore https://kedacore.github.io/charts
helm repo add podinfo https://stefanprodan.github.io/podinfo

Update the Helm Repos

helm repo update

Install KEDA

Create a namespace and install the Keda Helm chart in it

kubectl create namespace keda-demo
helm install keda kedacore/keda --namespace keda-demo --version 2.3.2

Install the Metrics Exporter Workloads (PodInfo)

Install the PodInfo helm chart in the same namespace as Keda. You want to use the --set flag to enable the ServiceMonitor. If the prometheus-operator is configured correctly then the ServiceMonitor will be detected and Prometheus will be configured to scape metrics from the PodInfo workload(s).

helm install podinfo --namespace keda-demo podinfo/podinfo --version 5.2.1 --set serviceMonitor.enabled=true

After installation, go to the /prometheus/targets endpoint for you prometheus instance and confirm that the PodInfo workload is a target that is successfully being scraped.

Install a Deployment

Deploy a Deployment object keda-demo namespace.

kubectl apply -n keda-demo -f examples/deployments/example-workload.yaml

Demos

Autoscaling a Deployment with a Prometheus Metrics

Prerequisites

• Have the prometheus-operator installed and configured to look for ServiceMonitors in the keda-demo namespace. You can use the community helm chart to install it
• Optional, have Ambassador's Telepresence v2 installed

Steps

1. Update the serverAddress in the examples/keda/prom-scaledobject.yaml file.

   • If you want to access it interally, you can use the format http://SVC_NAME.NAMESPACE.svc.cluster.local:9090. If your prometheus instance is served on a subpath you must attach the subpath after the port number (e.g. http://SVC_NAME.NAMESPACE.svc.cluster.local:9090/prometheus)

2. Deploy the ScaledObject with the prometheus trigger

   kubectl apply -f examples/keda/prom-scaledobject.yaml
   

3. Check the scaled object is ready with kubectl get scaledobject -n keda-demo

4. If it is not ready, check the logs of the keda pod (keda logs POD_NAME -n keda-demo). It'll most likely be the server address being incorrect.

5. Check and examine the HPA object created by KEDA with kubectl get hpa -n keda-demo and kubectl describe hpa -n keda-demo.

6. Open another terminal and watch the pods in the keda-demo namespace with kubectl get pods -n keda-demo -w.

7. Open another terminal and watch the deployments in the keda-demo namespace with kubectl get deployments -n keda-demo -w.

8. Now we want to trigger the autoscaling. If you have telepresence, repeatedly run curl commands against the PodInfo service with while :; do curl podinfo.keda-demo:9898; sleep 1; done.

   • If you don't have telepresence you can exec into another pod with curl installed and run curl podinfo.keda-demo.svc.cluster.local:9898/metrics from within the context of that pod,

9. As you run the curl commands, you'll eventually notice the number of replicas of the consumer workload increase in the other terminals.

10. Stop running the curl commands.

11. Wait until the Deployment scales down back to the minimum number of replicas

12. Clean up the resources.

    kubectl delete -f examples/keda/prom-scaledobject.yaml
    

Autoscaling a Deployment with a Redis List

Steps

1. Deploy the redis Deployment and Service

   kubectl apply -f examples/deployments/redis.yaml
   kubectl apply -f examples/deployments/redis-svc.yaml
   

2. Confirm the redis pod is running with kubectl get pods -n keda-demo

3. Confirm the redis service has successfully found the redis pod by checking an endpoint exists with kubectl get endpoints -n keda-demo

4. Deploy the ScaledObject with the redis trigger

   kubectl apply -f examples/keda/redis-scaledobject.yaml
   

5. Check the scaled object is ready with kubectl get scaledobject -n keda-demo

6. If it is not ready, check the logs of the keda pod (keda logs POD_NAME -n keda-demo). It'll most likely be the server address being incorrect.

7. Check and examine the HPA object created by KEDA with kubectl get hpa -n keda-demo and kubectl describe hpa -n keda-demo.

8. Open another terminal and watch the pods in the keda-demo namespace with kubectl get pods -n keda-demo -w

9. Open another terminal and watch the deployments in the keda-demo namespace with kubectl get deployments -n keda-demo -w

10. Now we want to trigger the autoscaling. Exec into the redis pod with kubectl exec redis-(RANDOM_STRING) -it -n keda-demo -- redis-cli

11. Check the length of the list stored in the mylist key with LLEN mylist

12. Add to the list with LLPUSH mylist "string" until the length of the list is above the threshold.

13. One additional replica should have been created.

14. Remove items from the list with LPOP mylist until the length of the list is below the threshold.

15. Wait until the Deployment scales down back to the minimum number of replicas.

16. Clean up the resources.

      kubectl delete -f examples/deployments/redis.yaml
      kubectl delete -f examples/deployments/redis-svc.yaml
      kubectl delete -f examples/keda/redis-scaledobject.yaml
    

Autoscaling Jobs with a Redis List

Steps

1. Deploy the redis Deployment and Service

   kubectl apply -f examples/deployments/redis.yaml
   kubectl apply -f examples/deployments/redis-svc.yaml
   

2. Confirm the redis pod is running with kubectl get pods -n keda-demo

3. Confirm the redis service has successfully found the redis pod by checking an endpoint exists with kubectl get endpoints -n keda-demo

4. Deploy the ScaledJob with the redis trigger.

   kubectl apply -f examples/keda/redis-scaledjob.yaml
   

5. Check the scaled object is ready with kubectl get scaledjob -n keda-demo

6. If it is not ready, check the logs of the keda pod (keda logs POD_NAME -n keda-demo). It'll most likely be the server address being incorrect.

7. Open another terminal and watch the pods in the keda-demo namespace with kubectl get pods -n keda-demo -w

8. Open another terminal and watch the jobs in the keda-demo namespace with kubectl get jobs -n keda-demo -w

9. Now we want to trigger the autoscaling. Exec into the redis pod with kubectl exec redis-(RANDOM_STRING) -it -n keda-demo -- redis-cli

10. Check the length of the list stored in the myotherlist key with LLEN myotherlist

11. Add to the list with LLPUSH myotherlist "myotherlist" until the length of the list is above the threshold.

12. One additional replica should have been created.

13. Wait and observe as more jobs continously get created.

14. Remove the item from the list with LPOP myotherlist

15. Wait and observe as no more jobs are created.

16. Clean up the resources.

      kubectl delete -f examples/deployments/redis.yaml
      kubectl delete -f examples/deployments/redis-svc.yaml
      kubectl delete -f examples/keda/redis-scaledjob.yaml