Migration From Function app to AKS

Migration of functions running on function app to AKS Cluster Using Cronjob for cost optimization

Nityanand Kumar's photo
Nityanand Kumar
·

2 min read

  1. Need to Analyze Cost Difference Between Two Azure Services

    We need to calculate the total cost of all functions running on a Function App and estimate the probable cost of running the same functions on an AKS node. This analysis will help understand the cost difference between the two setups.

  2. Design Cronjob Manifests

    • Choose Cronjob Triggers: Map your Timer or scheduled triggers from the Function App to Kubernetes Cronjobs.

    • Write Cronjob YAML: Create a Kubernetes Cronjob YAML file for each function. Example:

        yamlCopyEditapiVersion: batch/v1
  kind: CronJob
  metadata:
    name: my-function-cronjob
  spec:
    schedule: "*/5 * * * *" # Every 5 minutes
    jobTemplate:
      spec:
        template:
          spec:
            containers:
            - name: my-function
              image: <registry-name>/<image-name>:<tag>
              resources:
                limits:
                  memory: "512Mi"
                  cpu: "500m"
            restartPolicy: OnFailure

3. Deploy to AKS

  • Apply the Cronjob YAML to your AKS cluster: You can apply this file directly from local but best practice is to deploy using standard git model

      kubectl apply -f <cronjob-file>.yaml

  • Verify the CronJob:

      kubectl get cronjob
  kubectl get pods

4. Configure Logging and Monitoring

  • Enable Azure Monitor for AKS to track resource usage and performance.

  • Monitor Container performance using Container insights

  • Set up Prometheus and Grafana for detailed metrics if needed.

  • Ensure logs from your containers are sent to Azure Log Analytics or another logging solution.

5. Test Your Functions

  • Validate the execution of each Cronjob.

  • Check if the containerized functions behave as expected.

  • Confirm resource utilization matches predictions.

6. Optimize Costs

  • Review AKS node sizing to ensure efficient resource allocation.

  • Use Kubernetes autoscaling:

    • Cluster Autoscaler: Scales AKS nodes.

    • Horizontal Pod Autoscaler (HPA): Scales pods based on CPU/memory.

  • Monitor resource usage over time and adjust CronJob schedules or node pools as needed.

7. Decommission the Function App

  • Once all functions are running successfully on AKS, decommission the original Function App to avoid redundant costs.

8. Continuously Monitor and Improve

  • Periodically review:

    • Cronjobs schedules and resource limits.

    • AKS cluster costs and scaling behavior.

    • Overall performance and cost savings.

Azurecost-optimisationaksFunction appscronjob