Autoscaling Kubernetes Workloads Based On Metrics
summary
Shows how to wire Prometheus through the Keptn Metrics Operator into Kubernetes so the HorizontalPodAutoscaler can use custom metrics; demo reads a Prometheus value of 2.84 with a target of 1 and automatically scales a deployment from 1 to the configured max of 3 replicas. Explains CRDs for metric providers, a configurable fetch interval, and that this approach exposes metrics via the Kubernetes API or Prometheus but does not auto-scale nodes.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:48Manual scaling challenges
- 4:19Kubernetes autoscalers overview
- 7:52Why HPA is powerful
- 8:56Observability tools issues
- 10:02Introducing Captain Matrix operator
- 11:06Demo setup prerequisites
- 12:12Deploying app and Prometheus
- 14:01Configuring Captain Matrix provider
- 15:29HPA with Captain Matrix metrics
- 16:16Live demo of HPA
- 17:21Benefits of Captain Matrix operator
- 17:43How Captain Matrix operator works
- 18:24Advanced use cases
- 21:18Q&A
related talks
34mAutoscale your Kubernetes Workload with Prometheus - Frederic Branczyk, CoreOS
30mHorizontal Pod Autoscaler Reloaded - Scale on Custom Metrics
51mHorizontal Autoscaling with Kubernetes
33mDeep Dive: Kubernetes Metric APIs using Prometheus
35mAutoscaling Kubernetes Deployments: A (Mostly) Practical Guide
19mKeynote: Monitoring Kubernetes Clusters with Prometheus
29mKubernetes Autoscaling: Beyond One Size Fits All
13mUsing Virtual Nodes to Auto-Scale Your Application on Kubernetes