Probes

Liveness (health) check

Liveness (health) of Hive Router is available at /health endpoint, and represents the HTTP health status of the router.

A request to this endpoint returns 200 OK if the router's HTTP layer is healthy. Health requests pass through plugin on_http_request hooks and their on_end callbacks, so a plugin can explicitly change or end the response. Coprocessors do not run for health requests.

If a container fails its liveness probe repeatedly, the runtime environment should restart or replace the container with a new one.

To manually check your router's health status, you can send an HTTP GET request to the /health endpoint. For example, using curl:

curl -I http://localhost:4000/health

Readiness check

Readiness of Hive Router is available at /readiness endpoint, and represents the readiness status and ability to process GraphQL requests.

The readiness request passes through plugin on_http_request hooks and their on_end callbacks. The router first uses a supergraph selected by a plugin for that request, then falls back to the configured supergraph. It returns 200 OK only when the selected supergraph has not retired and its router runtime is usable. Missing selection or runtime construction failure returns 503 Service Unavailable. Coprocessors do not run for readiness requests.

With supergraph.source: plugin, there is no configured fallback. The plugin must call set_supergraph during every readiness request. This selection is request-local and does not configure a default for later requests.

With a configured source, a router that has not loaded its first supergraph returns 503 Service Unavailable. If the initial supergraph loaded successfully but a later reload fails, the previous usable generation remains selected and readiness continues to report 200 OK. This prevents the runtime from replacing a working instance because a new generation failed to load.

To manually check your router's readiness status, you can send an HTTP GET request to the /readiness endpoint. For example, using curl:

# Replace the URL with your Router's host and port
curl -I http://localhost:4000/readiness

Prometheus endpoint

Prometheus requests continue to bypass both plugins and coprocessors. Unlike health and readiness, a plugin cannot select a request-local supergraph or change the Prometheus response.