Kubernetes Objects Consume Hardware Resources

Nodes can use memory, disk, and cpu.
Pods can use the same.

Kubernetes Objects Consume Hardware Resources

Scheduler Considers these hardware resources

Kubernetes schedulers "decide" which node that a pod gets allocated to. The Scheduler considers resources that are required by a pod. The scheduler also considers resources "left available" on the host(s) environmnets.

Resource Maximums Can Be Configured

According to the k8s docs on configuring memory limits, A Kubernetes cluster can be divided into namespaces. Once you have a namespace that has a default memory limit, and you then try to create a Pod with a container that does not specify its own memory limit, then the control plane assigns the default memory limit to that container.

Resource Limits can Be describe in LimitRange definition files

Memory

Here is a memory LimitRange example for a pod. K8s Memory units are describe bytes- integers with quantity suffixes of:

E
P
T
G for gigabytes (1,000,000,000 bytes)
M for Megabytes (1,000,000 bytes)
k for kilobytes (1,000 bytes)
Ei
Pi
Ti
Gi for gibibyte (1,073,741,824 bytes)
Mi for Mebibyte (1,048,576 bytes)
Ki for Kibibyte (1,024 bytes)

in This def file is described

requesting 256MiB of memory
setting a max mem limit of 512 MiB

# mem-limits.yaml
apiVersion: v1
kind: LimitRange
metadata:
  name: memory-limit-range
spec:
  limits:
    - default:
        memory: 512Mi
      defaultRequest:
        memory: 256Mi
      type: Container

This can be applied to a namespace for all pods in the namespace, here the default namespace. With this applied, any container created within a pod in the default namespace, where the pod itself does not specify its own memory min or max, the k8s control plane applies the described values to the pod.

kubectl apply -f mem-limits.yaml --namespace=default

CPU

Here is a CPU LimitRange example for a pod. K8s CPU units of measurement are equivelant to 1 physical cpu core, or 1 virtual core - depending on the host env type. The below def file describes:

requesting 256MiB of memory
setting a max mem limit of 512 MiB

# cpu-limits.yaml
apiVersion: v1
kind: LimitRange
metadata:
  name: memory-limit-range
spec:
  limits:
    - default:
        cpu: 1
      defaultRequest:
        cpu: 0.5
      type: Container

This can be applied to a namespace for all pods in the namespace, here the default namespace. With this applied, any container created within a pod in the default namespace, where the pod itself does not specify its own cpu min or max, the k8s control plane applies the described values to the pod.

kubectl apply -f cpu-limits.yaml --namespace=default

Resources Can Be Configured by Pod Definition

Here, an example of a pod that includes explicity memory & cpu request sizes. here,

2 containers, both with similar resource descriptions
- requested ammounts
  - .25 CPU units
  - 64MiB of memory
- max limits
  - .5 CPU units
  - 128MiB of memory

apiVersion: v1
kind: Pod
metadata:
  name: frontend
spec:
  containers:
    - name: api-app
      image: some-api-image:v2
      # here!
      resources:
        requests:
          memory: "64Mi"
          cpu: "250m"
        limits:
          memory: "128Mi"
          cpu: "500m"
    - name: hidden-worker
      image: some-worker-image:v2
      # here!
      resources:
        requests:
          memory: "64Mi"
          cpu: "250m"
        limits:
          memory: "128Mi"
          cpu: "500m"

Understanding Resource Allocation on Running Pods

kubectl get pod pod-name-here -o yaml

This command will reveal containers:-resource:(limits + request) details.

Pod Resource Requests and Kubernetes Scheduler Impacts

Pods get created.
K8s scheduler selects a node for the pod to run on.
Each node, itself, has a maximum capacity of resources that the node can allocate to pods.
The scheduler asserts that the maximum available resources of the node are always larger than the sum of the resource requests of the containers (inside pods).
Resource Limit Description Without Request Definition Pod resource limits can be described without describing the resource request.

Tags:

Kubernetes