Install Avuru Obs and discover your first service map
Start on a test cluster. eBPF discovers supported traffic without application changes; OpenTelemetry extends collection where you need it. Startup depends on your cluster, image downloads and application traffic.
1. Check prerequisites
You need Helm, a reachable Kubernetes context, persistent storage for ClickHouse and Linux nodes compatible with the sensor. Read the Kubernetes requirements. eBPF collection needs node-level permissions; check your cluster's security constraints.
2. Install
# Flagship install — the chart is published to GHCR as an OCI artifact,
# eBPF auto-discovers your services. No repo to add.
helm install avuruobs oci://ghcr.io/avuruvision/charts/avuruobs \
--version 0.16.0 -n avuruobs --create-namespace
# Point apps at the gateway (OTLP):
# http://avuruobs-gateway:4318 (HTTP)
# http://avuruobs-gateway:4317 (gRPC)
# Open the UI:
kubectl -n avuruobs port-forward svc/avuruobs-ui 8080:80
The chart brings together the sensor, gateway, hub, UI and ClickHouse. The example uses a published release. Docker supports local evaluation; the VM path uses your OTLP pipeline.
3. Open the UI
kubectl -n avuruobs port-forward svc/avuruobs-ui 8080:80
Open http://localhost:8080/ and select the receiving project. Generate traffic to a supported application: an empty map on an idle cluster is not necessarily a collection failure.
4. Verify
kubectl -n avuruobs get pods
Check that the components are ready, then find a service on the map and a recent trace. If ClickHouse is pending, inspect its PVC and StorageClass. If signals are missing, check sensor permissions and logs, kernel/protocol support and the selected project.
5. Send your existing OTLP
From a different Kubernetes namespace, use the complete service DNS name:
export OTEL_EXPORTER_OTLP_TRACES_ENDPOINT=http://avuruobs-gateway.avuruobs.svc.cluster.local:4318/v1/traces
This endpoint is for OTLP/HTTP. gRPC uses port 4317 with the corresponding SDK configuration. Read the OTLP guide for ingest keys, dual export and other signals.
Next
- Understand the service map.
- Investigate a slow checkout with a reproducible synthetic example.
- Choose your modules and check their prerequisites.