How long before the deployment scales back to zero replica as there are zero incoming request

[jaiprasad266]

The httpscaledobject configuration is like below:

The first two request to curl -H "Host: myhost.com" 192.168.5.15/query returned error on backend (dial tcp 10.43.149.72:8080: connect: connection refused) , and third request was successful and the replica got scaled up to 1.

With zero incoming request , it took more than 8 minutes to scale back to 0 .
I was expecting it scale down bit faster.

kind: HTTPScaledObject
apiVersion: http.keda.sh/v1alpha1
metadata:
    name: {{ include "quarkus-rest-server.fullname" . }}
spec:
    host: myhost.com
    targetPendingRequests: 100
    scaleTargetRef:
        deployment: quarkus-rest-server
        service: quarkus-rest-server
        port: 8080
    replicas:
        min: 0
        max: 5    

[nrukavkov]

try to play with timeouts like this

helm install http-add-on kedacore/keda-add-ons-http --namespace keda --set interceptor.tcpConnectTimeout=30000ms --set interceptor.responseHeaderTimeout=30000ms --set interceptor.tlsHandshakeTimeout=30s

[jaiprasad266]

sure thanks , will try this and post an update

[pschulten]

tried it. doesn't make a difference. My scale-down to zero time is 5m.
I guess it's the default cooldownPeriod of the underlying ScaledObject https://github.com/kedacore/keda-docs/blob/82fb80bd2c2c7291d0a6abff6e17b5c1903b1f1c/content/docs/2.0/concepts/scaling-deployments.md?plain=1#L48

it's (as of now) not configurable :(

http-add-on/operator/controllers/scaled_object.go

Line 29 in bbe8abb

appScaledObject, appErr := k8s.NewScaledObject(