Skip to main content

​
Concept

Every node has its own capacity of resources (CPU and memory). When you deploy a pod, it consumes the node’s resources. The kube-scheduler is responsible for scheduling pods to nodes. It identifies the node with the required resources to deploy the pod. If a node does not have the required resources, the pod will be in pending state. 
# See pod events — will show "insufficient resources" if pending
kubectl describe pods <pod-name>

​
Resource Requests

Resource requests are the minimum amount of CPU or memory guaranteed to a container. By default, every pod has a default request of 0.5 CPU and 256MB of memory. CPU units:
  • 0.1 CPU = 100m (milli)
  • 1 CPU = 1 vCPU (AWS) = 1 Core (GCP, Azure) = 1 Hyperthread
Memory units:
BinaryBytes
1G (Gigabyte)1,000,000,000
1M (Megabyte)1,000,000
1K (Kilobyte)1,000
1Gi (Gibibyte)1,073,741,824
1Mi (Mebibyte)1,048,576
1Ki (Kibibyte)1,024
pod.yaml
apiVersion: v1
kind: Pod
metadata:
  name: sample-pod
spec:
  containers:
    - name: sample-container
      image: ubuntu
      resources:
        requests:
          memory: "2Gi"
          cpu: 2

​
Resource Limits

Resource limits are the maximum amount of CPU or memory a container can use. By default, every pod has a default limit of 1 CPU and 512MB of memory. If a pod exceeds its limits:
  • CPU: The system throttles the CPU so it does not go beyond the limit.
  • Memory: The system kills the container with an OOM (Out of Memory) error.
pod.yaml
apiVersion: v1
kind: Pod
metadata:
  name: sample-pod
spec:
  containers:
    - name: sample-container
      image: ubuntu
      resources:
        requests:
          memory: "2Gi"
          cpu: 2
        limits:
          memory: "4Gi"
          cpu: 4

​
Resource Quota

See Resource Quota for Namespace for setting total resource limits at the namespace level.

​
Limit Range

See Limit Range for Namespace for setting per-container default, minimum, and maximum resource limits.

Build docs developers (and LLMs) love

Get started for freeTalk to us