Topic 02: ReplicaSets Link to heading

A ReplicaSet ensures a specified number of pod replicas are running at all times. If a pod dies, the ReplicaSet creates a replacement. In practice you rarely create ReplicaSets directly — Deployments manage them for you. But you need to know the structure because Deployments use the same pattern.

YAML Link to heading

apiVersion: apps/v1
kind: ReplicaSet
metadata:
  name: web-rs
spec:
  replicas: 3
  selector:
    matchLabels:
      app: web # which pods this RS manages
  template:
    metadata:
      labels:
        app: web # must match selector — RS stamps these on pods it creates
    spec:
      containers:
      - name: web
        image: nginx

The Three Label Locations Link to heading

ReplicaSets have labels in three distinct places:

metadata:
  labels:
    app: web # labels ON the RS object itself — for filtering the RS
spec:
  selector:
    matchLabels:
      app: web # the query — "find pods with these labels"
  template:
    metadata:
      labels:
        app: web # labels stamped on pods — must match selector
  • metadata.labels — cosmetic, used to filter the RS object itself
  • spec.selector.matchLabels — the query the RS uses to find its pods
  • spec.template.metadata.labels — what gets stamped on pods at creation

If selector and template labels don’t match, the API rejects the resource at creation.

Imperative Link to heading

# No direct imperative for ReplicaSet — generate YAML
k create deployment web --image=nginx $do # then change kind to ReplicaSet

# Scale
k scale rs web-rs --replicas=5

# Delete RS but keep pods
k delete rs web-rs --cascade=orphan

k get rs
k describe rs web-rs

Common Mistakes Link to heading

  • selector.matchLabels and template.metadata.labels must match exactly — mismatch = API rejection
  • ReplicaSet will adopt any existing pod in the namespace that matches its selector — even pods it didn’t create

Exam Tips Link to heading

  • In practice: always use Deployments, not ReplicaSets directly
  • k get rs — check ReplicaSet status after a deployment rollout
  • ReplicaSet controller does NOT retry after FailedCreate for a missing ServiceAccount — must delete and recreate the Deployment