KEDA should report it state/availability to the kubernetes cluster.
Duplicate of https://github.com/kedacore/keda/issues/237?
@tomkerkhove no, these are probes for the KEDA Operator deployment, to expose KEDA state to the cluster. #237 is talking about probes on the target deployment and scaling based on the probes there.
Oh I've missed that, fully agree on this one then!
Fixed in #792
Can we get some docs on this on https://keda.sh/docs/2.0/operate/?
I am not sure this is something worth documenting. Those probes exposes state of the Operator and Metrics Server Deployment toward kubernetes, eg. kubernetes scheduler knows when the Deployment is not available and should try to spin another pod.
Let's close since it's part of our YAML & chart?
@zroubalik @tomkerkhove Where can I see the Yamls and Chart? I want to see examples of liveness/readiness configuration.
Chart can be found here: https://github.com/kedacore/charts/tree/master/keda
@tomkerkhove I don't see there an example for scaledobject.yaml.
This is what I have from version 1.5.0. Should it tweaked to be compatible with v2.0.0?
apiVersion: keda.k8s.io/v1alpha1
kind: ScaledObject
metadata:
name: lf-scaledobject-{{ .Values.tenantId }}
namespace: {{ .Values.lf_namespace }}
labels:
deploymentName: log-forwarding-{{ .Values.tenantId }}
spec:
scaleTargetRef:
deploymentName: log-forwarding-{{ .Values.tenantId }}
pollingInterval: 180
cooldownPeriod: 43200
minReplicaCount: 0
maxReplicaCount: 5
triggers:
- type: gcp-pubsub
metadata:
subscriptionSize: "50"
subscriptionName: "lf-data-{{ .Values.tenantId }}-sub"
credentials: GENERIC_GOOGLE_CREDENTIALS
The probes are for KEDA runtime, ScaledObject scales your workload, so you are in control of those probes.
Or maybe I'm not fully getting your question.
You can refer Migration Guide https://keda.sh/docs/2.0/migration/
Exactly what I needed :) Thanks!