OTel Labs Platform – OpenTelemetry on EKS: End-to-End Observability

Infrastructure and GitOps platform for the otel-labs OpenTelemetry demo application.

The repository provisions an Amazon EKS cluster using Terraform and deploys platform components such as ArgoCD, Traefik, OpenTelemetry Collector, and Kubernetes observability tooling using GitOps.

The companion application repository is: ◉ otel-labs

otel-labs contains the instrumented applications:

  • Node Frontend
  • Python Orders API
  • Go Inventory API

Container images built from otel-labs are consumed and deployed by this repository.

Architecture

Components

Infrastructure

Provisioned using Terraform:

  • VPC
  • Public Subnets
  • Amazon EKS
  • Managed Node Group
  • EKS Addons

GitOps

Managed using ArgoCD:

  • Traefik
  • OpenTelemetry Collector
  • kube-state-metrics
  • Node Frontend
  • Python Orders
  • Go Inventory

Observability

OpenTelemetry Collector receives:

  • Metrics
  • Logs
  • Traces

and exports telemetry to:

  • New Relic
  • Honeycomb

The collector also gathers Kubernetes telemetry using:

  • kubeletstats
  • k8s_cluster
  • kube-state-metrics

Repository Structure

.
├── terraform/
│   ├── eks.tf
│   ├── nodegroup.tf
│   ├── addons.tf
│   ├── bootstrap.sh
│   └── ...
│
└── k8s/
    ├── argo-apps/
    ├── node-frontend/
    ├── python-orders/
    ├── go-inventory/
    ├── otel-collector/
    └── kustomization.yml

Relationship with otel-labs

This repository manages the platform.

The application repository manages the workloads.

When a new image is built from otel-labs, ArgoCD can deploy the updated image into the Kubernetes cluster managed by this repository.

Cluster Bootstrap

Provision the cluster:

cd terraform

terraform init

terraform plan

terraform apply

Terraform creates:

  • VPC
  • EKS Cluster
  • Managed Node Group
  • EKS Addons

After the infrastructure is ready, the bootstrap script installs ArgoCD and applies required Kubernetes resources.

Terraform Apply
       │
       ▼
Amazon EKS
       │
       ▼
bootstrap.sh
       │
       ├── Update kubeconfig
       ├── Install ArgoCD
       ├── Create Namespaces
       └── Apply Secrets

If you want, you can use kubectl apply -f root_apps.yaml in the bootstartp so the workload is also created automatically

Deploying ArgoCD Applications

The repository supports three deployment approaches.

Option 1: Individual Applications

Create each ArgoCD Application separately.

kubectl apply -f k8s/argo-apps/traefik-app.yml

kubectl apply -f k8s/argo-apps/kube-state-metrics-app.yml

kubectl apply -f k8s/argo-apps/otel-collector-app.yml

kubectl apply -f k8s/argo-apps/node-frontend-app.yml

kubectl apply -f k8s/argo-apps/python-orders-app.yml

kubectl apply -f k8s/argo-apps/go-inventory-app.yml

Option 2: Root Application (App of Apps)

A single parent application creates all child applications.

Root Application
      │
      ├── Traefik
      ├── kube-state-metrics
      ├── OpenTelemetry Collector
      ├── Node Frontend
      ├── Python Orders
      └── Go Inventory

This is the classic ArgoCD App of Apps pattern.

Option 3: Kustomize Bootstrap

Apply the root Kustomization.

kubectl apply -k k8s

k8s/
│
├── argo-apps/
│   ├── *.yml
│
└── kustomization.yml

This creates all ArgoCD Applications in a single command.

Accessing ArgoCD

Port-forward locally:

kubectl port-forward svc/argocd-server -n argocd 8080:80

Or access through the configured Traefik ingress by applying the ingress

kubectl apply -fk8s/infra/argo/argo-ingress.yml

Kubernetes Observability

The OpenTelemetry Collector gathers cluster telemetry from multiple sources.

kubeletstats
      │
      ├── Node Metrics
      ├── Pod Metrics
      └── Container Metrics

k8s_cluster
      │
      ├── Cluster State
      └── Kubernetes Events

kube-state-metrics
      │
      ├── Deployments
      ├── StatefulSets
      ├── Services
      └── Replica Status

This telemetry is exported together with application telemetry to both observability backends.

Claude MCP

Integrate Honeycomb MCP server with Claude to and ask questions in natural language about cluster and applications. Honeycomb canvas also has the same AI capabilities.

DevSecOps Implementation