Kubeadapt

Quick Start

This guide walks you through installing Kubeadapt in your Kubernetes cluster using Helm.

Prerequisites

Before installing Kubeadapt, ensure you have:

  • Kubernetes Cluster - Version 1.24 or later
  • Helm 3 - Version 3.0 or later
  • kubectl - Configured to access your cluster
  • Storage Class – Required for Prometheus data persistence. Since Kubeadapt only requires a maximum 30-minute data retention policy for Prometheus, you may use an ephemeral emptyDir volume instead of persistent storage.

Installation

Step 1: Add Helm Repository

bash
1helm repo add kubeadapt https://kubeadapt.github.io/kubeadapt-helm 2helm repo update

Step 2: Create Namespace

bash
1kubectl create namespace kubeadapt

Step 3: Generate Agent Token

Before installation, generate an agent token from the Kubeadapt dashboard:

  1. Navigate to app.kubeadapt.io
  2. Go to Clusters → Add Cluster
  3. Choose the cloud provider or custom cluster type
  4. Specify environment type production-like or non-production-like
  5. If you already have a running Node Exporter in your cluster, it can be re-used. There is no need to install it again. You can optionally provide the namespace where it is installed.

The generated command will look similar to this:

bash
1helm install kubeadapt kubeadapt/kubeadapt \ 2 --namespace kubeadapt \ 3 --create-namespace \ 4 --set agent.enabled=true \ 5 --set agent.config.token="your-generated-token-here"

Note: The actual command will be automatically generated with your unique token. Copy and use the command provided in the dashboard.

Step 4: Install Kubeadapt

Use the dashboard-generated command from Step 3 above, which includes your agent token.

Step 5: Verify Installation

Check all pods are running:

bash
1kubectl get pods -n kubeadapt

Expected output:

text
1NAME READY STATUS AGE 2kubeadapt-kube-state-metrics-xxxxx 1/1 Running 2m 3kubeadapt-opencost-xxxxx 1/1 Running 2m 4kubeadapt-prometheus-node-exporter-xxxxx 1/1 Running 2m 5kubeadapt-prometheus-server-xxxxx 1/1 Running 2m

Check services:

bash
1kubectl get svc -n kubeadapt

Expected output:

text
1NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE 2kubeadapt-prometheus-server ClusterIP 10.0.0.100 <none> 80/TCP,9090/TCP 2m 3kubeadapt-opencost ClusterIP 10.0.0.101 <none> 9003/TCP 2m 4kubeadapt-kube-state-metrics ClusterIP 10.0.0.102 <none> 8080/TCP 2m 5kubeadapt-prometheus-node-exporter ClusterIP 10.0.0.103 <none> 9100/TCP 2m

Configuration Options

Storage Configuration

Persistent Storage (any StorageClass):

yaml
1prometheus: 2 server: 3 persistentVolume: 4 enabled: true 5 size: 20Gi 6 storageClass: your-storage-class # gp2, gp3, standard, ssd, managed-premium, local-path, etc. 7 retention: "30m"

Ephemeral Storage (emptyDir):

yaml
1prometheus: 2 server: 3 persistentVolume: 4 enabled: false 5 emptyDir: 6 enabled: true 7 sizeLimit: "20Gi" 8 retention: "30m"

Resource Configuration

Adjust resources based on cluster size.

Small Clusters (~10 nodes, ~300 pods):

yaml
1agent: 2 enabled: true 3 resources: 4 requests: 5 cpu: 100m 6 memory: 128Mi 7 limits: 8 cpu: 500m 9 memory: 256Mi 10 11prometheus: 12 server: 13 persistentVolume: 14 enabled: true 15 size: 10Gi # With networkCost: ~10Gi (~10K-180K network time series) 16 storageClass: gp2 # Adjust for your cloud provider 17 retention: "30m" 18 resources: 19 requests: 20 cpu: 500m # With networkCost: ~575m (+15% scraping overhead) 21 memory: 512Mi # With networkCost: ~1-2Gi (network time series vary by topology) 22 limits: 23 cpu: 2000m # With networkCost: ~2500m (burst capacity) 24 memory: 2Gi # With networkCost: ~3-4Gi (traffic bursts) 25 26 kube-state-metrics: 27 resources: 28 requests: 29 cpu: 10m 30 memory: 55Mi 31 limits: 32 cpu: 100m 33 memory: 128Mi 34 35 prometheus-node-exporter: 36 resources: 37 requests: 38 cpu: 100m 39 memory: 30Mi 40 limits: 41 cpu: 200m 42 memory: 50Mi 43 44opencost: 45 opencost: 46 exporter: 47 resources: 48 requests: 49 cpu: 10m 50 memory: 55Mi 51 limits: 52 cpu: 999m 53 memory: 1Gi

Medium Clusters (~100 nodes, ~3,000 pods):

yaml
1agent: 2 enabled: true 3 resources: 4 requests: 5 cpu: 200m 6 memory: 256Mi 7 limits: 8 cpu: 1000m 9 memory: 1Gi 10 11prometheus: 12 server: 13 persistentVolume: 14 enabled: true 15 size: 20Gi # With networkCost: ~30Gi (~100K realistic, ~18M theoretical network time series) 16 storageClass: gp2 17 retention: "30m" 18 resources: 19 requests: 20 cpu: 1000m # With networkCost: ~1150m (+15% scraping overhead) 21 memory: 2Gi # With networkCost: ~3-6Gi (network topology dependent) 22 limits: 23 cpu: 4000m # With networkCost: ~5000m (burst capacity) 24 memory: 8Gi # With networkCost: ~12Gi (high-density service mesh) 25 26 kube-state-metrics: 27 resources: 28 requests: 29 cpu: 100m 30 memory: 256Mi 31 limits: 32 cpu: 500m 33 memory: 1Gi 34 35 prometheus-node-exporter: 36 resources: 37 requests: 38 cpu: 100m 39 memory: 30Mi 40 limits: 41 cpu: 200m 42 memory: 50Mi 43 44opencost: 45 opencost: 46 exporter: 47 resources: 48 requests: 49 cpu: 50m 50 memory: 128Mi 51 limits: 52 cpu: 1500m 53 memory: 2Gi

Large Clusters (~500 nodes, ~15,000 pods):

yaml
1agent: 2 enabled: true 3 resources: 4 requests: 5 cpu: 500m 6 memory: 512Mi 7 limits: 8 cpu: 3000m 9 memory: 4Gi 10 config: 11 queryConcurrency: 20 12 goMaxProcs: 6 13 goMemLimit: "3600MiB" 14 15prometheus: 16 server: 17 persistentVolume: 18 enabled: true 19 size: 50Gi # With networkCost: ~100Gi 20 storageClass: gp2 21 retention: "30m" 22 resources: 23 requests: 24 cpu: 2000m # With networkCost: ~2500m (+15% scraping overhead) 25 memory: 8Gi # With networkCost: ~12-24Gi (network topology dependent) 26 limits: 27 cpu: 8000m # With networkCost: ~12000m (burst capacity) 28 memory: 24Gi # With networkCost: ~32Gi (high-density service mesh) 29 30 kube-state-metrics: 31 resources: 32 requests: 33 cpu: 200m 34 memory: 512Mi 35 limits: 36 cpu: 1000m 37 memory: 2Gi 38 39 prometheus-node-exporter: 40 resources: 41 requests: 42 cpu: 100m 43 memory: 30Mi 44 limits: 45 cpu: 200m 46 memory: 50Mi 47 48opencost: 49 opencost: 50 exporter: 51 resources: 52 requests: 53 cpu: 200m 54 memory: 512Mi 55 limits: 56 cpu: 2000m 57 memory: 4Gi

Data Retention Configuration

Adjust based on your monitoring needs:

yaml
1prometheus: 2 server: 3 retention: "30m" 4opencost: 5 opencost: 6 dataRetention: 7 dailyResolutionDays: 15

Note: Metrics are continuously sent to Kubeadapt cloud for long-term storage and analysis. Local retention primarily serves as a buffer for the agent's 60-second collection cycle.

Upgrading

Dashboard-Managed Upgrades (Recommended)

Kubeadapt Dashboard automatically tracks your cluster's Helm chart version and notifies you when upgrades are available.

How it works:

  1. Version Tracking: Dashboard monitors each cluster's current agent version via heartbeat
  2. Upgrade Detection: Uses semantic versioning (semVer) to detect available updates
  3. One-Click Command: Dashboard generates an upgrade command that preserves your custom values

To upgrade:

  1. Navigate to app.kubeadapt.io → Your Cluster → Connectivity tab
  2. If an update is available, you'll see an "Update Available" badge
  3. Click Copy Upgrade Command to get the optimized command
  4. Run the command in your terminal

Example upgrade command:

bash
1helm get values kubeadapt -n kubeadapt > values.yaml && \ 2helm upgrade kubeadapt kubeadapt/kubeadapt \ 3 --version 0.5.3 \ 4 --namespace kubeadapt \ 5 -f values.yaml

Why this approach:

  • Preserves Configuration: Exports your current values before upgrading
  • Version-Specific: Targets exact version tested by Kubeadapt team for current installed version
  • Breaking Changes: Dashboard warns if upgrade contains breaking changes

Manual Upgrade

If you prefer manual upgrades:

bash
1# Update Helm repository 2helm repo update 3 4# Export current values 5helm get values kubeadapt -n kubeadapt > values.yaml 6 7# Upgrade to latest version 8helm upgrade kubeadapt kubeadapt/kubeadapt \ 9 --namespace kubeadapt \ 10 -f values.yaml 11 12# Or upgrade to specific version 13helm upgrade kubeadapt kubeadapt/kubeadapt \ 14 --version 0.5.3 \ 15 --namespace kubeadapt \ 16 -f values.yaml

Check Available Versions

bash
1helm search repo kubeadapt --versions

Uninstalling

Remove Kubeadapt:

bash
1helm uninstall kubeadapt --namespace kubeadapt

Delete namespace and all data:

bash
1kubectl delete namespace kubeadapt

Warning: This permanently deletes all Prometheus metrics and cost data.

Next Steps

After successful installation:

  1. Configure Cloud Integration - Set up AWS, GCP, or Azure for accurate billing costs
  2. Review Cost Attribution - Understand how costs are calculated
  3. Explore Features - Check out the Dashboard and Available Savings
  4. Implement Rightsizing - Follow the Rightsizing Guide to start optimizing workloads
Back to Docs
Edit this page