kubectl-autoscale

Auto scale a deployment with no target CPU utilization specified

$ kubectl autoscale [[deploy|deployment]] [deployment_name] --min [min_replicas] --max [max_replicas]

$ kubectl autoscale [[deploy|deployment]] [deployment_name] --max [max_replicas] --cpu-percent [target_cpu]

kubectl autoscale deployment|rs|rc <name> --min=<min-pods> --max=<max-pods> [--cpu-percent=<%%>] [--resource=cpu] [options]

--cpu-percent int32
    Target average CPU utilization percentage (default 80)

--defaults string
    Default metrics like 'cpu=80%%,memory=80%%' for unmatched requests

--disable-scaling
    If true, disables autoscaling on the target

--horizontal-pod-autoscaler-config string
    Path to HPA config file for defaults

--limits-cpu string
    CPU limit (e.g., 200m, 1) for resource spec

--limits-memory string
    Memory limit (e.g., 200Mi, 1Gi) for resource spec

--max int32
    Maximum number of pods

--min int32
    Minimum number of pods

--name string
    Name of the autoscaler (defaults to deployment name)

--requests-cpu string
    CPU request (e.g., 200m, 1) for resource spec

--requests-memory string
    Memory request (e.g., 200Mi, 1Gi) for resource spec

--resource string
    Metric resource name (default 'cpu'); requires --resource-value

--resource-value string
    Target value for the resource metric (default 50)

--value int32
    Target percentage of requested value (default 50)

--dry-run=client|server|none
    Dry run without applying changes

The kubectl autoscale command creates a HorizontalPodAutoscaler (HPA) resource that automatically adjusts the number of replicas in a Deployment, ReplicaSet, or ReplicationController based on observed metrics like CPU or memory utilization.

It enables horizontal scaling to maintain performance under varying loads by increasing or decreasing pod counts within specified min/max limits. For example, you can scale a deployment when average CPU usage exceeds 80%.

Requires a metrics server (like Prometheus or metrics-server) for CPU/memory metrics. Supports resource metrics (CPU/memory), custom metrics, and external metrics. Updates existing HPA if one already exists for the target.

Ideal for production workloads needing elasticity without manual intervention.

Requires metrics-server or compatible provider for CPU/memory metrics.
HPA cannot scale below min or above max replicas.
Does not handle vertical scaling or node autoscaling.
Existing HPA updated, not recreated.

kubectl autoscale deployment nginx --min=2 --max=10 --cpu-percent=80
Creates HPA scaling nginx deployment between 2-10 pods at 80% CPU.

Deploy metrics-server: kubectl apply -f https://github.com/kubernetes-sigs/metrics-server/releases/latest/download/components.yaml
Ensure workloads have resource requests/limits set.

Introduced in Kubernetes v1.1 (2015) for basic CPU scaling; enhanced in v1.23+ with memory, custom/external metrics, and improved defaults.

kubectl scale(1), kubectl get(1), kubectl describe(1)