| sectionid | sectionclass | title | parent-id |
|---|---|---|---|
lab2-HPA |
h2 |
Autoscaling |
lab-clusterapp |
In this section we will explore how the Horizontal Pod Autoscaler (HPA) can be used and works within Kubernetes/OpenShift.
As defined in the Kubernetes documentation:
Horizontal Pod Autoscaler automatically scales the number of pods in a replication controller, deployment, replica set or stateful set based on observed CPU utilization.
We will create an HPA and then use OSToy to generate CPU intensive workloads. We will then observe how the HPA will scale up the number of pods in order to handle the increased workloads.
{% collapsible %}
Run the following command to create the autoscaler. This will create an HPA that maintains between 1 and 10 replicas of the Pods controlled by the ostoy-microservice DeploymentConfig created. Roughly speaking, the HPA will increase and decrease the number of replicas (via the deployment) to maintain an average CPU utilization across all Pods of 80% (since each pod requests 50 millicores, this means average CPU usage of 40 millicores)
oc autoscale deployment/ostoy-microservice --cpu-percent=80 --min=1 --max=10
In the OSToy app in the left menu click on "Autoscaling" to access this portion of the workshop.
As was in the networking section you will see the total number of pods available for the microservice by counting the number of colored boxes. In this case we have only one. This can be verified through the web UI or from the CLI.
You can use the following command to see the running microservice pods only:
oc get pods --field-selector=status.phase=Running | grep microservice
Now that we know that we only have one pod let's increase the workload that the pod needs to perform. Click the link in the center of the card that says "increase the load". Please click only ONCE!
This will generate some CPU intensive calculations. (If you are curious about what it is doing you can click here).
Note: The page may become slightly unresponsive. This is normal; so be patient while the new pods spin up.
After about a minute the new pods will show up on the page (represented by the colored rectangles). Confirm that the pods did indeed scale up through the OpenShift Web Console or the CLI (you can use the command above).
Note: The page may still lag a bit which is normal.
After confirming that the autoscaler did spin up new pods, revisit Azure Monitor like we did in the logging section. By clickin on the containers tab we can see the resource consumption of the pods and see that three pods were created to handle the load.
{% endcollapsible %}