These operational endpoints cover health checks, readiness, billing, and Prometheus metrics on the management API port (:8080). They are not gated by the readiness guard described on the Management API overview, so they are available as soon as the HTTP server starts.

GET /health

Returns the process health as plain text. This is the liveness signal.

Response: text/plain

Condition	HTTP status	Body
Healthy	200	`healthy`
Not healthy	500	`not healthy`

GET /health
→ 200 OK
healthy

Use this for a liveness probe or a health-indicator badge; poll every 5–10 seconds.

GET /ready

Returns the readiness state as a raw JSON HealthState object (no envelope). This is the Kubernetes readiness probe endpoint.

Response: application/json

Condition	HTTP status	Body
Ready	200	`HealthState` JSON
Not ready	503	`HealthState` JSON

A node that is not ready answers 503, whether it is still starting up or already running. A probe treats any non-2xx as "not ready"; a check that compares the status code exactly should expect 503.

{
  "is_healthy": true,
  "is_ready": true,
  "current_leadership_role": "leader"
}

Field	Type	Description
`is_healthy`	`boolean`	Whether the process is healthy
`is_ready`	`boolean`	Whether the node is ready to accept traffic
`current_leadership_role`	`string`	The cluster leadership role: `leader`, `follower`, `mixed` (leads some shards), or `standalone` for a single node
`store_ack_policy`	`string`	The next storage engine's acknowledgement policy, `strict` or `fast`; omitted on the legacy engine
`reason`	`string`	Why the node is not ready, when the cause is one to act on rather than wait out; omitted otherwise

Use this for a readiness probe and to display each node's cluster leadership role.

A clustered node also reports not ready when it has been unable to reach the cluster coordinator over the internal forwarding network for more than ten seconds. Without this check the node would look healthy while every consumer-group request sent to it waited minutes and failed. In that state reason says so and names the nodes it can still reach:

{
  "is_healthy": true,
  "is_ready": false,
  "current_leadership_role": "follower",
  "store_ack_policy": "fast",
  "reason": "this node has not reached the cluster coordinator over the internal forwarding network for 28s; nodes it can reach: kubemq-next-2:6800"
}

The node recovers on its own when the link returns. The check applies only to clustered nodes on the next storage engine, and never to the coordinator itself; a rolling restart does not trip it, because a node that goes away is replaced as coordinator rather than waited for.

Kubernetes probes

Wire /health and /ready to the container's liveness and readiness probes on the management API port (8080):

livenessProbe:
  httpGet:
    path: /health
    port: 8080
  initialDelaySeconds: 10
  periodSeconds: 10
readinessProbe:
  httpGet:
    path: /ready
    port: 8080
  initialDelaySeconds: 5
  periodSeconds: 5

For full deployment settings, see Kubernetes configuration.

GET /billing

Returns a usage summary for the current node, wrapped in the standard response envelope.

Response: standard envelope with Billing data.

Field	Type	Description
`hostname`	`string`	Node hostname
`uptime`	`float64`	Uptime in seconds
`messages`	`int64`	Total messages processed
`volume`	`float64`	Total data volume in bytes
`last_message`	`int64`	Unix timestamp of the last message

{
  "error": false,
  "error_string": "",
  "data": {
    "hostname": "kubemq-node-0",
    "uptime": 3600,
    "messages": 50000,
    "volume": 1024000,
    "last_message": 1709312400
  }
}

GET /metrics

Returns all registered metrics in Prometheus text exposition format. This is the endpoint Prometheus scrapes; it is not meant for direct UI consumption — use the snapshot endpoints for dashboard data instead.

Response: text/plain (Prometheus exposition format)

For the full metric catalog, labels, and PromQL examples, see Prometheus Metrics.

GET /kill

Disabled. Always returns HTTP 404 in plain text and does not use the standard envelope. Reserved — do not use.

GET /kill
→ 404 Not Found
{"message":"not found"}

Health & System Endpoints