Skip to content

Run Obstack Bee

Obstack Bee is a host-level Linux observability agent. It discovers application processes from the host’s /proc, captures TCP traffic with eBPF, parses application protocols, correlates calls where it has evidence, and exports traces as OTLP/HTTP protobuf.

This guide gets a demo from zero to a visible trace using the published obstackit/bee:0.3.0 image. It includes complete Docker Compose and Kubernetes definitions and two collector choices:

  • OpenTelemetry Collector prints traces and can forward them to any OTLP backend.
  • Vector prints traces and writes newline-delimited JSON for direct inspection.
application containers or pods
│ TCP observed at the host cgroup
▼
one Bee agent per Linux node
│ OTLP/HTTP protobuf :4318
▼
Vector or OpenTelemetry Collector
│
├── terminal / JSON file for this demo
└── your tracing backend
Environment Start here Collector supplied by the example
One Linux host, Docker, or Docker Swarm Docker demo Choose Vector or OpenTelemetry Collector
Kubernetes, k3s, or kind on Linux nodes Kubernetes demo Node-local OpenTelemetry Collector sidecar

Read requirements before starting. Bee needs host visibility and kernel privileges that ordinary application containers do not need.

After sending a request to the included bee-demo service:

  1. http://127.0.0.1:9400/livez returns ok.
  2. http://127.0.0.1:9400/readyz returns ready after the collector accepts telemetry.
  3. obstack_bee_ebpf_packets_total and obstack_bee_spans_exported_total increase.
  4. The collector prints or stores a SERVER span for service bee-demo.

Bee exports traces over OTLP. Its own health and pressure metrics use Prometheus exposition at /metrics; its current service graph is JSON at /graph.

The public image contains the production binary and requires a signed evaluation or subscription license. Ask for one by submitting an information request on www.obstack.it before starting the agent.