Kubernetes QoS Guaranteed class¶

Kubernetes QoS Guaranteed class is a Quality of Service (QoS) class assigned to Pods to ensure they receive a fixed amount of resources. This class represents the highest priority for resource scheduling and stability in Kubernetes^[400-devops__06-Kubernetes__k8s-ithelp__Day21__README.md].

Definition¶

A Pod is classified as Guaranteed when specific conditions are met regarding its CPU and memory resource requests and limits^[400-devops__06-Kubernetes__k8s-ithelp__Day21__README.md].

For a Pod to be categorized in this class, every container within it must satisfy the following equality^[400-devops__06-Kubernetes__k8s-ithelp__Day21__README.md]:

request.memory == limit.memory
request.cpu == limit.cpu

This means that the container is guaranteed a specific amount of resources and cannot consume more than that amount.

Configuration Example¶

The following manifest demonstrates a Pod configuration that results in a Guaranteed QoS class^[400-devops__06-Kubernetes__k8s-ithelp__Day21__README.md].

apiVersion: v1
kind: Pod
metadata:
  name: qos-demo
  namespace: qos-example
spec:
  containers:
  - name: qos-demo-ctr
    image: nginx
    resources:
      limits:
        memory: "200Mi"
        cpu: "700m"
      requests:
        memory: "200Mi"
        cpu: "700m"

Characteristics and Behavior¶

Priority and Eviction¶

Pods in the Guaranteed class are assigned the highest priority level compared to other QoS classes^[400-devops__06-Kubernetes__k8s-ithelp__Day21__README.md].

Stability: They are typically protected from being killed or throttled as long as their resource usage remains within the defined limits^[400-devops__06-Kubernetes__k8s-ithelp__Day21__README.md].
Eviction Order: When system memory is insufficient, Kubernetes relies on QoS classes to determine eviction order. Best-Effort Pods (lowest priority) are evicted first. Guaranteed Pods are the last to be evicted, generally only killed if they exceed their resource limits and no lower-priority Pods are available to terminate^[400-devops__06-Kubernetes__k8s-ithelp__Day21__README.md].

Resource Comparison¶

The Guaranteed class contrasts with other QoS tiers in how resources are handled:

vs. Burstable: Burstable Pods have a minimum resource guarantee but are allowed to exceed their requests up to their limits. Guaranteed Pods operate strictly at a fixed capacity^[400-devops__06-Kubernetes__k8s-ithelp__Day21__README.md].
vs. BestEffort: BestEffort Pods have no resource requests or limits and can use available resources sparingly, but they are the first to be terminated under resource pressure^[400-devops__06-Kubernetes__k8s-ithelp__Day21__README.md].

Sources¶

400-devops__06-Kubernetes__k8s-ithelp__Day21__README.md