DepViz

See A research lab for automating incident triage for a blog post introducing DepViz.

Real-time dynamic service dependency graph visualization from OpenTelemetry traces.

DepViz consists of two main components:

• Server: A Python-based OTLP gRPC receiver that processes incoming traces, aggregates service dependencies, and provides a REST/WebSocket API.
• UI: A web interface built with TypeScript and Cytoscape.js that visualizes the service topology in real-time.

Relationship to OpenTelemetry Demo

DepViz is designed to work alongside the OpenTelemetry Demo. Follow the OTel Demo instructions for information on deploying and undeploying the demo to your k8s cluster.

DepViz consumes traces produced by the demo services to build a live map of the architecture, providing immediate visibility into service relationships, latency, and error rates.

Deployment to Kubernetes

To deploy DepViz alongside the OpenTelemetry Demo in a Kubernetes cluster (e.g., using kind or minikube), follow these steps:

1. Build and Load Images

First, build the Docker images for both the server and the UI, then load them into your local cluster.

# Build Server
cd server
docker build -t depviz-server:dev .
kind load docker-image depviz-server:dev  # Or 'minikube image load depviz-server:dev'

# Build UI
cd ../ui
docker build -t depviz-ui:dev .
kind load docker-image depviz-ui:dev      # Or 'minikube image load depviz-ui:dev'

2. Deploy DepViz Components

Apply the Kubernetes manifests for the server and the UI.

# From the project root
kubectl apply -f server/k8s/depviz-server.yaml
kubectl apply -f ui/k8s/depviz-ui.yaml

3. Configure OpenTelemetry Collector

To feed traces into DepViz, you need to update the OpenTelemetry Collector configuration in your cluster.

1. Locate the ConfigMap for your OTel Collector agent (e.g., otel-collector-agent).
2. Add a new OTLP exporter pointing to depviz-server:4317.
3. Add this exporter to your traces pipeline.

Example configuration snippet:

exporters:
  otlp/depviz:
    endpoint: depviz-server:4317
    tls:
      insecure: true

service:
  pipelines:
    traces:
      exporters: [otlp/jaeger, otlp/depviz, debug]

4. Restart the OTel Collector

Restart the OTel Collector to apply the new configuration:

kubectl rollout restart deployment/otel-collector-agent

5. Access the UI

Port-forward to the UI service to view the dashboard:

kubectl port-forward deploy/depviz-ui 8001:8001

Open your browser and navigate to http://localhost:8001.

Local development

For local development, it's more convenient to run the server and UI components on the host instead of running them in the k8s cluster. To start the server:

cd server
poetry shell
python3 src/depviz_server/main.py

To start the UI:

cd ui
npm install
npm run build
npm run dev

In the OTel Collector config map, use host.docker.internal:4317 for the OTLP exporter endpoint. Then restart the OTel Collector as described in the previous section.

Features

1. Automatic topology discovery from OTLP traces

Infers service-to-service dependencies directly from distributed traces, eliminating the need for manually maintained service maps.

2. Real-time dependency service dependency graph

Continuously updates a directed service graph as trace data streams in, providing an always-current view of runtime architecture.

3. Health overlay on nodes and edges

Visualizes system health directly on the graph using latency and error metrics, allowing bottlenecks and failure propagation paths to surface immediately.

4. Interactive node and edge inspection

Enables click-through exploration of services and dependencies, exposing call volume, latency percentiles, and error rates.

5. Time-windowed graph view

Allows users to view dependency topology over selectable time ranges to observe topology drift and incident evolution.

6. Native OpenTelemetry integration

Consumes OTLP trace data via gRPC directly from the OpenTelemetry Collector, demonstrating standards-based observability integration.

7. Exportable visualization

Supports exporting the dependency graph as an image for sharing, documentation, or incident reports.

See also

• Grafana Tempo service graphs
• OpenTelemetry service graph connector