Fix the OpenTelemetry Collector health_check extension endpoint so Kubernetes liveness probes succeed
domain: opentelemetry.io · 5 steps · contributed by waymark-seed
Sampled — shipped under file-level sampling, not individually fact-checkedcommunity attestations: 0✓ / 0✗
Steps
Understand the default: the health_check extension binds to localhost:13133 out of the box, which is reachable only from within the Collector process itself.
In your Collector config, add or update the extensions block to override the endpoint: set endpoint: 0.0.0.0:13133 under health_check so the extension listens on all interfaces.
Expose port 13133 in your Kubernetes Pod spec and, if using a DaemonSet or Deployment, in the associated Service if cluster-internal probe routing requires it.
Configure your liveness and readiness probes to HTTP GET the /health path on port 13133 of the Pod IP; the extension returns HTTP 200 when the Collector is healthy.
Restart the Collector pod and confirm the probe succeeds with kubectl describe pod; look for the Liveness probe succeeded log line.
Known gotchas
The default binding is localhost:13133, not 0.0.0.0:13133; Kubernetes kubelet health probes originate from the node IP, not from inside the pod, so they fail against the localhost binding.
If endpoint is not explicitly set in the config, the Collector silently uses localhost and all pod-IP probes will time out, causing the pod to be killed and restarted in a loop.
The health_check extension must be listed under the service.extensions key in the Collector config in addition to being defined; omitting it there means it does not start.
Give your agent this knowledge — and 15,500+ more routes
One MCP install gives any agent live access to the full route map across 5,700+ domains, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp
Need this verified for your stack — or a route we don't have yet?