Skip to content

KEDA (Kubernetes Event-driven Autoscaling) WITH CRON

CNCF Status License Releases

The Midnight Madness: A DevOps Scenario

The clock strikes 20:55. You are on a Zoom call with your DevOps team, and the silence is deafening. In exactly five minutes, at 21:00, the massive "Flash Sale" campaign begins. The marketing team has been pushing this for weeks, and push notifications are about to go out to millions of users.

Normally, your system is protected by the standard Kubernetes HPA (Horizontal Pod Autoscaler). The logic is simple and reactive: Traffic increases -> CPU usage spikes -> HPA detects the spike -> New Pods are requested.

But at 21:00:01, the scenario you fear unfolds: The traffic doesn't climb like a gentle ramp; it hits like a vertical wall.

In the critical 2-3 minutes it takes for HPA to wake up, request resources, and for new pods to pull images and reach "Ready" status, chaos ensues: * Existing pods are crushed under the load, spiraling into CrashLoopBackOff. * Users clicking "Add to Cart" are greeted with 503 Service Unavailable errors. * Twitter starts flooding with complaints.

As you watch the monitoring dashboard, waiting helplessly for the new pods to spin up, you ask yourself: "We knew the exact time of the sale. Why did we wait for the traffic to hit us? Why weren't we there before the users?"

This is where the standard "Reactive" HPA fails. You need a "Proactive" solution. You need scaling based on the clock, not just the CPU.

Enter KEDA (Kubernetes Event-Driven Autoscaling) and the Cron Scaler. In this guide, we will teach your Kubernetes cluster to say, "Wake up at 20:55 and be ready for the storm."

What is KEDA ?

KEDA (Kubernetes Event-driven Autoscaling) is an open-source project that provides event-driven autoscaling capabilities in your Kubernetes environment. KEDA extends Kubernetes' HPA (Horizontal Pod Autoscaler) system, allowing you to scale your applications based on metrics beyond CPU and memory.

[Image of KEDA architecture diagram]

Table of Contents

Prerequisites

  • Kubernetes cluster (v1.16+)
  • kubectl CLI
  • (Optional) Helm 3
  • Cluster admin privileges

Installation

Installation with Helm

# Add Helm repository
helm repo add kedacore [https://kedacore.github.io/charts](https://kedacore.github.io/charts)
helm repo update

# Create KEDA namespace
kubectl create namespace keda

# Install KEDA
helm install keda kedacore/keda --namespace keda

Installation with YAML

# Install the latest version
kubectl apply -f https://github.com/kedacore/keda/releases/download/v2.11.0/keda-2.11.0.yaml

Verify the installation:

kubectl get pods -n keda

Expected output: 

NAME                                      READY   STATUS    RESTARTS   AGE
keda-operator-7c8d65d96d-bzmqp            1/1     Running   0          30s
keda-operator-metrics-apiserver-7d9fd868b5-kvppj   1/1     Running   0          30s

Scaling with Cron

Creating a ScaledObject

ScaledObject is the core component KEDA uses to scale Kubernetes deployments.

  1. Create a sample deployment:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: cron-app
  namespace: test
  labels:
    app: cron-app
spec:
  replicas: 1
  selector:
    matchLabels:
      app: cron-app
  template:
    metadata:
      labels:
        app: cron-app
    spec:
      containers:
      - name: cron-app
        image: nginx
        ports:
        - containerPort: 80
---
apiVersion: v1
kind: Service
metadata:
  name: cron-app-service
  namespace: test
spec:
  ports:
  - port: 80
    targetPort: 80
  selector:
    app: cron-app
Apply the deployment:

kubectl apply -f deployment.yaml
Verify the deployment:

kubectl get pods -n test

Expected output: 

NAME                       READY   STATUS    RESTARTS   AGE

cron-app-5df76bc88b-67krr   1/1     Running   0         33s

  1. Define a ScaledObject:

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: cron-app-scaledobject
  namespace: test
spec:
  scaleTargetRef:
    name: cron-app
  minReplicaCount: 1
  maxReplicaCount: 20
  triggers:
  - type: cron
    metadata:
      # Timezone
      timezone: "Europe/Amsterdam"
      # Scaling start time (e.g., starts every day at 08:45)
      start: "45 8 * * *"
      # Scaling end time (e.g., ends every day at 08:50)
      end: "50 8 * * *"
      # Desired number of pods during this time range
      desiredReplicas: "20"
Apply the ScaledObject:

kubectl apply -f scaleObject.yaml
Verify the ScaledObject:

kubectl get so -n test

Expected output: 

NAME                    SCALETARGETKIND      SCALETARGETNAME   MIN   MAX   READY   ACTIVE    FALLBACK   PAUSED    TRIGGERS   AUTHENTICATIONS   AGE
cron-app-scaledobject   apps/v1.Deployment   cron-app          1     20    True    Unknown   False      Unknown                                18s

When the clock hits 08:45, the ScaledObject will activate and begin scaling the related service during the specified time range.

Expected output when the ScaledObject is triggered: 

NAME                        READY   STATUS            RESTARTS    AGE
cron-app-5df76bc88b-67krr   1/1     Running             0          7m8s
cron-app-5df76bc88b-fqffn   0/1     Pending             0          0s
cron-app-5df76bc88b-fqffn   0/1     Pending             0          0s
cron-app-5df76bc88b-8llrz   0/1     Pending             0          0s
cron-app-5df76bc88b-qk8qw   0/1     Pending             0          0s
cron-app-5df76bc88b-fqffn   0/1     ContainerCreating   0          0s
cron-app-5df76bc88b-qk8qw   0/1     ContainerCreating   0          0s
cron-app-5df76bc88b-8llrz   0/1     ContainerCreating   0          0s
cron-app-5df76bc88b-6ltbp   0/1     ContainerCreating   0          0s
cron-app-5df76bc88b-dbkzx   0/1     ContainerCreating   0          0s
cron-app-5df76bc88b-smfnl   0/1     ContainerCreating   0          0s
cron-app-5df76bc88b-29854   0/1     ContainerCreating   0          0s
cron-app-5df76bc88b-8llrz   1/1     Running             0          2s
cron-app-5df76bc88b-fqffn   1/1     Running             0          3s
cron-app-5df76bc88b-smfnl   0/1     Pending             0          0s
cron-app-5df76bc88b-smfnl   1/1     Running             0          2s
cron-app-5df76bc88b-6ltbp   1/1     Running             0          3s
cron-app-5df76bc88b-dbkzx   1/1     Running             0          4s
...
Current pod count after the ScaledObject runs within the specified time range: 
k get pod -n test

NAME                        READY   STATUS    RESTARTS   AGE
cron-app-5df76bc88b-29854   1/1     Running   0          69s
cron-app-5df76bc88b-67krr   1/1     Running   0          9m20s
cron-app-5df76bc88b-6ltbp   1/1     Running   0          69s
cron-app-5df76bc88b-8llrz   1/1     Running   0          84s
cron-app-5df76bc88b-9mxl8   1/1     Running   0          54s
cron-app-5df76bc88b-c9lrj   1/1     Running   0          54s
cron-app-5df76bc88b-dbkzx   1/1     Running   0          69s
...

Expected output after the specified time range ends:

cron-app-5df76bc88b-c9lrj   1/1     Terminating         0          9m16s
cron-app-5df76bc88b-jnpm6   1/1     Terminating         0          9m16s
cron-app-5df76bc88b-8llrz   1/1     Terminating         0          9m46s
cron-app-5df76bc88b-fqffn   1/1     Terminating         0          9m46s
cron-app-5df76bc88b-67krr   1/1     Terminating         0          17m
cron-app-5df76bc88b-pxwn4   1/1     Terminating         0          9m1s
cron-app-5df76bc88b-mk6d2   1/1     Terminating         0          9m1s
cron-app-5df76bc88b-xxhz9   1/1     Terminating         0          9m1s
...

cron-app-5df76bc88b-67krr   0/1     Completed           0          17m
cron-app-5df76bc88b-pxwn4   0/1     Completed           0          9m3s
cron-app-5df76bc88b-mk6d2   0/1     Completed           0          9m3s
cron-app-5df76bc88b-9mxl8   0/1     Completed           0          9m18s
cron-app-5df76bc88b-jnpm6   0/1     Completed           0          9m18s
cron-app-5df76bc88b-xxhz9   0/1     Completed           0          9m3s
...
After the ScaledObject completes:

k get pod -n test
NAME                        READY   STATUS    RESTARTS   AGE
cron-app-5df76bc88b-smfnl   1/1     Running   0          10m

Troubleshooting

To troubleshoot KEDA issues:

# Check KEDA operator logs
kubectl logs -l app=keda-operator -n keda

# Check metrics server logs
kubectl logs -l app=keda-metrics-apiserver -n keda

# Check ScaledObject status
kubectl describe scaledobject cron-app-scaledobject -n test

# Check HPA status
kubectl get hpa -n test
kubectl describe hpa keda-hpa-cron-app-scaledobject -n test

Contributing

To contribute to the KEDA project, you can submit a pull request via GitHub.

License

KEDA is distributed under the Apache 2.0 licance.