CPU Latest
Scale applications based on cpu metrics.
Notice:
- This scaler requires prerequisites. See the ‘Prerequisites’ section.
- This scaler can scale to 0 only when user defines at least one additional scaler which is not CPU or Memory (eg. Kafka + CPU, or Prometheus + CPU) and
minReplicaCount
is 0.- This scaler only applies to ScaledObject, not to Scaling Jobs.
Prerequisites
KEDA uses standard cpu
and memory
metrics from the Kubernetes Metrics Server, which is not installed by default on certain Kubernetes deployments such as EKS on AWS. Additionally, the resources
section of the relevant Kubernetes Pods must include at least one of requests
or limits
.
- The Kubernetes Metrics Server must be installed. Installation instructions vary based on your Kubernetes provider.
- The configuration for your Kubernetes Pods must include a
resources
section with specifiedrequests
(orlimits
). See Resource Management for Pods and Containers. If the resources section is empty (resources: {}
or similar), KEDA checks if thedefaultRequest
(ordefault
for limits) is set inLimitRange
for theContainer
type in the same namespace. IfdefaultRequest
(ordefault
for limits) is missing too, the errormissing request for {cpu/memory}
occurs.
# a working example of resources with specified requests
spec:
containers:
- name: app
image: images.my-company.example/app:v4
resources:
requests:
memory: "128Mi"
cpu: "500m"
Trigger Specification
This specification describes the cpu
trigger that scales based on cpu metrics.
triggers:
- type: cpu
metricType: Utilization # Allowed types are 'Utilization' or 'AverageValue'
metadata:
type: Utilization # Deprecated in favor of trigger.metricType; allowed types are 'Utilization' or 'AverageValue'
value: "60"
containerName: "" # Optional. You can use this to target a specific container in a pod
Parameter list:
type
- Type of metric to use. Options areUtilization
, orAverageValue
.value
- Value to trigger scaling actions for:- When using
Utilization
, the target value is the average of the resource metric across all relevant pods, represented as a percentage of the requested value of the resource for the pods. - When using
AverageValue
, the target value is the target value of the average of the metric across all relevant pods (quantity).
- When using
containerName
- Name of the specific container to scale based on its CPU, rather than the entire pod. Defaults to empty if not specified.
💡 NOTE:
containerName
parameter requires Kubernetes cluster version 1.20 or higher withHPAContainerMetrics
feature enabled. Please see container resource metrics for more information.
💡 NOTE: The
type
parameter is deprecated in favor of the globalmetricType
and will be removed in a future release. Users are advised to usemetricType
instead.
Example
The following example targets CPU utilization of entire pod. If the pod has multiple containers, it will be sum of all the containers in it.
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
name: cpu-scaledobject
namespace: default
spec:
scaleTargetRef:
name: my-deployment
triggers:
- type: cpu
metricType: Utilization # Allowed types are 'Utilization' or 'AverageValue'
metadata:
value: "50"
The following example targets CPU utilization of a specific container (foo
) in a pod.
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
name: cpu-scaledobject
namespace: default
spec:
scaleTargetRef:
name: my-deployment
triggers:
- type: cpu
metricType: Utilization # Allowed types are 'Utilization' or 'AverageValue'
metadata:
value: "50"
containerName: "foo"