Skip to content

Keda Module Footprint ​

The Keda module consists of four workloads. Two of them, keda-manager and keda-operator, implement the Kubernetes operator pattern and consume extra resources only when they detect changes in the resources they watch, for example, Keda CR, ScaledObject CR, etc. Usually, they are idle and consume as little as few millicores of CPU time and less than 30MB of memory. At the time of active reconciliation, the observed CPU time jumps to 5m.

Similarly to the operators, the keda-admission-webhooks workload stays idle most of the time and performs validation operations only when you submit a new object from *.keda.sh API group.

The last workload, keda-operator-metrics-apiserver, continuously serves metrics for the Kubernetes autoscaling components. Here, the consumption is the highest, but in the case of one or two active KEDA scalers, it stays at 5-7 millicores of CPU time.

NameCPU (cores)Memory (bytes)
keda-admission-webhooks1m10Mi
keda-manager3m23Mi
keda-operator3m26Mi
keda-operator-metrics-apiserver5m30Mi