Pod autoscaling with custom-metrics

How can you use your custom-metrics that are in your monitoring system to scale your pods.

Transcript

1. Pod autoscaling with custom-metrics
2. Mattias Gees • Cloud Engineer @skyscrapers • Avid runner • Likes to experiment github.com/mattiasgees twitter.com/mattiasgees blog.mattiasgees.be
3. We are hiring! https://skyscrapers.eu/jobs
4. Who has used kubernetes?
5. API
6. Some history! • SIG Instrumentation • Tackled Heapster • Metrics-server (beta 1.8) • Annotations (1.2 - 1.5) • Blogposts Q1 2017 are outdated! • Autoscaler (1.6 - ...)
7. Architecture
8. Requirements • API aggregation layer • Resource Metrics API • Custom metrics API • kube-controller-manager • --horizontal-pod-autoscaler-use-rest-clients • --kubeconfig <path-to-kubeconfig> OR --master <ip- address-of-apiserver>
9. Enabled on 1.9
10. Basic check kubectl api-versions kubectl cluster-info
11. Architecture
12. Prometheus helm install coreos/prometheus-operator \ --name prometheus-operator \ --namespace monitoring helm install coreos/kube-prometheus \ --name kube-prometheus \ --namespace monitoring \ -f prometheus.yaml
13. Application kubectl apply -f metrics-app.yaml
14. Prometheus-adapter • Create cert • Upload cert as secret kubectl -n custom-metrics create secret tls cm- adapter-serving-certs --cert=server.crt -- key=server-key.pem • Deploy prometheus-adapter kubectl apply -f manifests
15. Test kubectl api-versions | grep custom custom.metrics.k8s.io/v1beta1
16. Test kubectl get --raw "/apis/custom.metrics.k8s.io/v1beta1/" | jq
17. Test kubectl get --raw \ "/apis/custom.metrics.k8s.io/v1beta1/namespaces/default/ pods/*/http_requests" \ | jq kubectl get --raw \ "/apis/custom.metrics.k8s.io/v1beta1/namespaces/default/ service/sample-metrics-app/http_requests" \ | jq
18. Deploy HPA kubectl apply -f hpa.yaml
19. Inspect kubectl describe hpa sample-metrics-app
20. Future • Still Alpha • Lots of discussions • Proposals for extra features
21. Questions?