Metrics server
- Docker & Kubernetes course
- Advanced
- Lesson 41/70
- 4 min
Learn that the metrics server is required for HPA.
Metrics server
The HPA gets CPU and memory metrics from metrics-server; metrics-server in turn collects them from the kubelet on each node. If it isn't installed, the HPA can do nothing and kubectl top doesn't work:
$ kubectl top pods
error: Metrics API not available
minikube addons enable metrics-server
# or
kubectl apply -f https://github.com/kubernetes-sigs/metrics-server/releases/latest/download/components.yaml
$ kubectl top pods -l app=api
NAME CPU(cores) MEMORY(bytes)
api-7f8c9d6b5-abcde 142m 180MiTesting under load
To see the HPA, artificial load is generated and watched with -w. Because of the metric collection and decision intervals, the reaction can take a minute or two:
kubectl run yuk --image=busybox:1.36 --restart=Never -- \
sh -c "while true; do wget -q -O- http://api:8000/; done"
kubectl get hpa api -w
NAME TARGETS REPLICAS
api cpu: 38%/70% 2
api cpu: 156%/70% 2
api cpu: 156%/70% 5
api cpu: 64%/70% 5
kubectl delete pod yukPractice A load test
Write three commands: enable the metrics-server add-on in Minikube; show the resource usage of the pods labelled app=api; watch the api HPA live.
Starter code
# buraya yazSample solution
minikube addons enable metrics-server
kubectl top pods -l app=api
kubectl get hpa api -wExpected output
🌟 The 'metrics-server' addon is enabled NAME CPU(cores) MEMORY(bytes) api-7f8c9d6b5-abcde 312m 190Mi NAME REFERENCE TARGETS MINPODS MAXPODS REPLICAS api Deployment/api cpu: 156%/70% 2 10 2 api Deployment/api cpu: 156%/70% 2 10 5
Questions and answers
Review the lesson questions together with their answers. To test yourself, solve the lesson interactively.
1 What is needed to scale on application metrics like request count?
- It can't be scaled
- A custom metrics provider like the Prometheus Adapter or KEDA Correct answer
- Only metrics-server
- Bigger nodes
Explanation: metrics-server only provides CPU and memory.
2 Show the pods' current resource usage.
kubectl ____ pods -l app=apikubectl top pods -l app=apiExplanation: kubectl top reads metrics-server data.
3 Match.
| Term | Meaning |
|---|---|
metrics-server |
CPU/memory metrics |
Metrics API not available |
No metrics-server |
KEDA |
Event-driven scaling |
hpa -w |
Watch the scaling |