K8s HPA (horizontal pod autoscaling)

Hi all,

I am trying to implement HPA in K8s.

I have set hpa with

kubectl autoscale deployment lms --cpu-percent=80 --min=1 --max=10

But it doesn’t work:

% kubectl describe hpa -n capnet-prod
Name:                                                  lms
Namespace:                                             <my namespace>
Labels:                                                <none>
Annotations:                                           <none>
CreationTimestamp:                                     Fri, 08 Jul 2022 05:44:12 -0300
Reference:                                             Deployment/lms
Metrics:                                               ( current / target )
  resource cpu on pods  (as a percentage of request):  <unknown> / 80%
Min replicas:                                          1
Max replicas:                                          10
Deployment pods:                                       1 current / 0 desired
Conditions:
  Type           Status  Reason                   Message
  ----           ------  ------                   -------
  AbleToScale    True    SucceededGetScale        the HPA controller was able to get the target's current scale
  ScalingActive  False   FailedGetResourceMetric  the HPA was unable to compute the replica count: failed to get cpu utilization: missing request for cpu

The reason seems to be that the pod specification does not include a request for cpu.

My questions are:

  • Is it possible to add a cpu request using a plugin? I don’t see a patch there to modify the deployment.
  • Which would be a good request for CPU? 500m sounds good?

If the answer to the first one is no, I would raise a PR with the following change to the lms deployment to something like:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: lms
  labels:
    app.kubernetes.io/name: lms
spec:
  template:
    spec:
      containers:
        - name: lms
...
          resources:
            requests:
              memory: 2Gi
{% if LMS_CPU_REQUEST is defined %}
              cpu: {{ LMS_CPU_REQUEST }}
{% endif %}

Similarly it can be repeated for cms and the workers.

What do you think?

A PR that adds the patch k8s-override was merged recently. Now you can leverage the patching functionality of kustomize to modify your deployments.

The folks at OpenCraft are using it to implement HPA.

2 Likes

Thanks @MoisesGonzalezS for your tips!

Based on all this information, I’ve created a simple plugin tutor-contrib-hpa to implement HPA. It’s highly inspired in OpenCraft’s Grove, but limited to only implement HPA.

3 Likes

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.