ReplicaSets and Self-Healing

A ReplicaSet has a replica count, a label selector and a Pod template whose labels match it. Its controller counts matching Pods and creates or deletes Pods until the count is right:

A ReplicaSet replacing a deleted front-end PodShell

kubectl apply -f - >/dev/null <<'EOF'
apiVersion: apps/v1
kind: ReplicaSet
metadata: { name: web }
spec:
  replicas: 2
  selector: { matchLabels: { app: booknest, tier: web } }
  template:
    metadata: { labels: { app: booknest, tier: web } }
    spec: { containers: [{ name: nginx, image: localhost:33500/booknest-web:1.3 }] }
EOF
kubectl wait --for=condition=Ready pod -l tier=web >/dev/null
kubectl delete $(kubectl get pod -l tier=web -o name | head -1) --wait=false >/dev/null
sleep 3 && kubectl get pods -l tier=web && kubectl delete replicaset web

Output

NAME        READY   STATUS    RESTARTS   AGE
web-cscc9   1/1     Running   0          4s
web-fz26x   1/1     Running   0          3s
replicaset.apps "web" deleted from booknest namespace

The replacement has a new name and IP, which is why objects find Pods by label. Each Pod names its owner in metadata.ownerReferences, so deleting the ReplicaSet deleted its Pods. Editing a ReplicaSet's template never changes existing Pods, so you rarely write one yourself.

MENU

ReplicaSets and Self-Healing