SECTOR 03 // REFERENCE

grpc

A standard gRPC health-check probe over the grpc.health.v1 protocol.

gRPC health. Calls the standard grpc.health.v1 Health/Check RPC and reports the serving status, so a gRPC backend gets a typed probe instead of an exec escape hatch. The target is dialed lazily (no connection until the first check), so configuring it before the service is up is fine.

providers:
  api:
    source: grpc
    config:
      target: "localhost:50051"   # host:port
      # tls: false                # plaintext by default (the common local/CI case)

target is the host:port to dial. tls is false by default (plaintext, the common local/CI case); set it true to dial over TLS.

Verbs #

health (probe) #

Checks one service, or the whole server. A SERVING response passes; any other status (NOT_SERVING, UNKNOWN, SERVICE_UNKNOWN) is a probe failure that still surfaces the status, so health works as a steadyState gate and the assertion can read what it saw.

Returns the status string as the value (SERVING, NOT_SERVING, …). meta.status (string) repeats it and meta.rpcMs (int) is the RPC latency. output is the status string.

steadyState:
  - run: api.health                 # whole server must be SERVING

verify:
  - run: api.health
    with: "orders.v1.Orders"        # one named service
    as: h
  - run: assert
    with: { of: "${.outputs.h.value}", equals: SERVING }
    desc: "orders service is serving"