karpenter-optimizer

Karpenter Optimizer Helm Chart

This Helm chart deploys Karpenter Optimizer on a Kubernetes cluster. Karpenter Optimizer analyzes your cluster usage and provides cost-optimized NodePool recommendations.

Prerequisites

Kubernetes 1.20+
Helm 3.0+
Karpenter installed (optional, for NodePool analysis)
RBAC permissions for reading nodes, pods, and NodePools

Installation

Add the Helm repository

helm repo add karpenter-optimizer https://charts.karpenter-optimizer.io
helm repo update

Install the chart

helm install karpenter-optimizer karpenter-optimizer/karpenter-optimizer \
  --namespace karpenter-optimizer \
  --create-namespace

Install with custom values

With Ollama (Legacy):

helm install karpenter-optimizer karpenter-optimizer/karpenter-optimizer \
  --namespace karpenter-optimizer \
  --create-namespace \
  --set config.ollama.enabled=true \
  --set config.ollama.url=http://ollama:11434 \
  --set config.ollama.model=granite4:latest

With IRSA (IAM Roles for Service Accounts) on EKS:

helm install karpenter-optimizer karpenter-optimizer/karpenter-optimizer \
  --namespace karpenter-optimizer \
  --create-namespace \
  --set serviceAccount.annotations."eks\.amazonaws\.com/role-arn"=arn:aws:iam::ACCOUNT_ID:role/karpenter-optimizer-role \
  --set config.aws.region=us-east-1

See IRSA Setup Guide for detailed instructions.

With LiteLLM:

helm install karpenter-optimizer karpenter-optimizer/karpenter-optimizer \
  --namespace karpenter-optimizer \
  --create-namespace \
  --set config.llm.enabled=true \
  --set config.llm.provider=litellm \
  --set config.llm.url=http://litellm-service:4000 \
  --set config.llm.model=gpt-3.5-turbo \
  --set config.llm.apiKey=your-api-key

Configuration

LLM Provider Configuration

Karpenter Optimizer supports multiple LLM providers for AI-enhanced explanations:

LiteLLM (Recommended)

LiteLLM provides a unified interface to multiple LLM providers (OpenAI, Anthropic, Azure OpenAI, etc.):

config:
  llm:
    enabled: true
    provider: "litellm"
    url: "http://litellm-service:4000"
    model: "gpt-3.5-turbo"
    apiKey: "your-api-key"  # Optional, if LiteLLM requires authentication

Ollama (Legacy)

Ollama is still supported for backward compatibility:

config:
  ollama:
    enabled: true
    url: "http://ollama-service:11434"
    model: "granite4:latest"

Or use the new unified LLM configuration:

config:
  llm:
    enabled: true
    provider: "ollama"
    url: "http://ollama-service:11434"
    model: "granite4:latest"

Configuration Parameters

The following table lists the configurable parameters and their default values:

Parameter	Description	Default
`replicaCount`	Number of replicas	`1`
`image.repository`	Image repository	`ghcr.io/kaskol10/karpenter-optimizer`
`image.tag`	Image tag	`""` (uses appVersion)
`image.pullPolicy`	Image pull policy	`IfNotPresent`
`service.type`	Service type	`ClusterIP`
`service.port`	Service port	`8080`
`ingress.enabled`	Enable ingress	`false`
`config.llm.enabled`	Enable LLM integration	`false`
`config.llm.provider`	LLM provider (`ollama` or `litellm`)	`ollama`
`config.llm.url`	LLM service URL	`""`
`config.llm.model`	Model name	`""`
`config.llm.apiKey`	API key (for LiteLLM)	`""`
`config.ollama.enabled`	Enable Ollama integration (legacy)	`false`
`config.ollama.url`	Ollama URL (legacy)	`""`
`config.ollama.model`	Ollama model (legacy)	`granite4:latest`
`config.aws.region`	AWS region	`us-east-1`
`resources.limits.cpu`	CPU limit	`500m`
`resources.limits.memory`	Memory limit	`512Mi`
`resources.requests.cpu`	CPU request	`100m`
`resources.requests.memory`	Memory request	`128Mi`

Values

See values.yaml for all available configuration options.

Uninstallation

helm uninstall karpenter-optimizer --namespace karpenter-optimizer

Upgrading

helm repo update
helm upgrade karpenter-optimizer karpenter-optimizer/karpenter-optimizer \
  --namespace karpenter-optimizer

Troubleshooting

Check pod status

kubectl get pods -n karpenter-optimizer

View logs

kubectl logs -n karpenter-optimizer -l app.kubernetes.io/name=karpenter-optimizer

Check service

kubectl get svc -n karpenter-optimizer

RBAC

The chart creates a ClusterRole and ClusterRoleBinding with the following permissions:

Read nodes, pods, namespaces
Read Karpenter NodePools
Read events and PodDisruptionBudgets

Security

Runs as non-root user (UID 1000)
Read-only root filesystem
Dropped capabilities
Pod Security Context configured

Additional Resources

Support

For issues and questions:

This site is open source. Improve this page.