Every pod requires some resources (CPU, memory and disk space). Based on the pod’s resource requirements and the resource availability on each of the nodes, the scheduler decides which node to spawn the pod on.

<aside> 💡 If none of the nodes have the required resources, the scheduler keeps the pod in the pending state.

</aside>

CPU and Memory metrics

Resource Requests

We can specify the resource requirements for a container in the definition file. This will be used by the scheduler to find out which node has these resources available to spawn a new pod. If not specified explicitly, the scheduler assumes that the pod will require 0.5 CPU and 256 Mi memory.

Resource Limits

A container has no limit on resources it can consume. A container can use up all the resources and starve the native processes of the host. However, pods support resource limits. They can be specified for each container in the pod definition file. If not specified explicitly, a container is given resource limits of 1 CPU and 512 Mi memory.

<aside> 💡 The container is not allowed to exceed the CPU limit. However, it can exceed the memory limit occasionally. If the container repeatedly exceeds its memory limit, k8s terminates the pod.

</aside>

Pod definition file

apiVersion: v1
kind: Pod
metadata:
	name: web-pod
spec:
	containers:
		- name: nginx
			image: nginx
			resources:
				requests:
					memory: "1Gi"
					cpu: 1
				limits:
					memory: "2Gi"
					cpu: 3