[health]

The /_gio/health endpoint load balancers and gio start poll: on by default, with a switch for the details it reports.

gio.toml

[health]
details = false     # only {"status":"ok","nodeReady":...}

Reference

Key	Default	Description
`enabled`boolean	`true`	Serve the endpoint. Turn it off when your platform probes another URL and you want no extra public endpoint.`0` / `false` / empty: `/_gio/health` answers `404`
`details`boolean	`true`	Report the deployment id, worker topology, cache size and uptime. Off, the answer keeps what a probe needs, so the deployment id and worker count are not public.`0` / `false` / empty: Only `status` and `nodeReady`

Behavior

The endpoint answers 200 from Rust while the process runs, without waiting for the worker:

json

{
  "cacheEntries": 0,
  "deploymentId": "495ddeb861e730d0",
  "http2": true,
  "nodeReady": true,
  "status": "ok",
  "tls": false,
  "uptimeSecs": 0,
  "workers": { "configured": 1, "ready": 1 }
}

nodeReady is false while no render worker is ready - during the respawn of the only worker, or of every worker in a pool. Cached and static content still serves, so the status stays 200: readiness probes should read the field.
With details = false the body is {"nodeReady":true,"status":"ok"}.
The endpoint is never rate-limited and no gio.toml or middleware.ts rule applies to it.

No key in this section logs a warning.

Examples

A Kubernetes readiness probe

The status is always 200, so use the HTTP check for liveness and read nodeReady for readiness (the app image has Node, so no extra tool is needed):

deployment.yaml

livenessProbe:
  httpGet: { path: /_gio/health, port: 3000 }
readinessProbe:
  exec:
    command:
      - node
      - -e
      - "fetch('http://127.0.0.1:3000/_gio/health').then(r => r.json()).then(h => process.exit(h.nodeReady ? 0 : 1), () => process.exit(1))"

Hide the deployment id