Question about spec.replicas.min and spec.replicas.max

[jesumyip]

What exactly do these 2 parameters control?

All the blogs and articles I have read do not explain it very well (or it's my complete lack of understanding that's preventing me from moving forward. :) )

kind: HTTPScaledObject
apiVersion: http.keda.sh/v1alpha1
metadata:
    name: test-test
spec:
    host: "mytest.test.com"
    targetPendingRequests: 10
    scaledownPeriod: 300
    scaleTargetRef:
        deployment: myapp
        service: myservice
        port: 8080
    replicas:
        min: 0
        max: 10

Assuming I have 11 pending requests in queue (assuming zero pods initially and I fire off curl 11 times), does this mean this add-on will spin up 1 pod in my deployment? If so, then when does the number of pods being spun up go from 1 to 2, 2 to 3, 3 to 4, until 9 to 10? And will it at any time jump by a factor larger than 1 i.e. 1 to 3 replicas (i.e. jump by a factor of 2 instead of 1)?

[tomkerkhove]

This is to control min/max replicas on the scale target, see https://keda.sh/docs/2.13/concepts/ to learn more

[jesumyip]

@tomkerkhove

When does the number of replicas go from 1 to 2 and then from 2 to 3?

[jesumyip]

Does it mean that for every 10 (I put 10 here because that is the value of spec.targetPendingRequests) pending requests inside the queue, Keda will add one more pod - up to the value of spec.replicas.max ?

So I can say total number of pods = int(number of pending requests in queue) / (spec.targetPendingRequests) (taking into account min and max replica values of course) ?

I am using the algorithm from https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale/#algorithm-details

[jesumyip]

I found the answer after more Googling. :)

https://github.com/kedacore/http-add-on/blob/main/docs/design.md#the-horizontal-pod-autoscaler