Requests, Limits & HPA

Requests, Limits & HPA

Step 1 of 12

replicas4average40%target60%

n4

n5

40

40

40

40

pend4

pend5

One bar per Pod: its CPU use as a percentage of its request, 250m. The HPA's target is 60%.

The Idea

A request is what the scheduler reserves for a container; a limit is the ceiling. Past its CPU limit a container is throttled; past its memory limit it can be OOM-killed. The HorizontalPodAutoscaler adds or removes Pods to hold a metric, such as CPU utilization measured against the request.

Real-World Example

An api Deployment requests 250m CPU per Pod and targets 60%. Traffic spikes, the average hits 90%, and the HPA computes ceil(4 × 90 / 60) = 6 replicas. If the new Pods do not fit, they wait Pending until node autoscaling adds a node.

The Tradeoff

Without a request, CPU utilization has no denominator and the HPA will not act on it. Requests set too high waste nodes; set too low, the scheduler packs Pods onto nodes that cannot serve them all. Scale-down waits out a stabilization window, so replicas do not flap.

Hands-On

# illustrative — requests and a limit on the container…
resources:
  requests: { cpu: 250m, memory: 256Mi }
  limits: { memory: 512Mi }
---
# …and an HPA that holds average CPU at 60% of the request
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata: { name: api }
spec:
  scaleTargetRef: { apiVersion: apps/v1, kind: Deployment, name: api }
  minReplicas: 4
  maxReplicas: 10
  metrics:
    - type: Resource
      resource: { name: cpu, target: { type: Utilization, averageUtilization: 60 } }

Your turn

Put the steps in the right order.

1. 1It computes ceil(4 × 90 / 60) = 6 and raises the Deployment's replicas↑↓
2. 2The metrics pipeline reports each Pod's CPU use↑↓
3. 3Two new Pods are scheduled, or wait Pending for a new node↑↓
4. 4The HPA compares average utilization with its 60% target↑↓

Check order

Mini quiz

1 / 3

Pods request 250m CPU and use 225m on average. The HPA sees CPU utilization of:

• A45%
• B90%
• C225%
• DNothing — it needs limits, not requests

Sources

• Resource management for Pods and containers — Kubernetes Documentation
• Horizontal Pod Autoscaling — Kubernetes Documentation
• Pod Quality of Service classes — Kubernetes Documentation
• Node autoscaling — Kubernetes Documentation