When something is wrong

Start with the first lines of the log, docker compose logs worker. Measured, on the path of Run Fruitback yourself:

[fruitback] listening on 0.0.0.0:8080 · store postgres · origins http://localhost:5500 · trusted proxy hops 0
[fruitback] who reads a site is chosen site by site, in the console: members only, or everyone. FRUITBACK_READ does not apply to a worker with accounts.
[fruitback] better-auth is ready, under /console/auth

The first line names the store, the origins and the hop count the worker actually runs with. A worker that is refused logs [fruitback] misconfigured, missing: … — /health will report 503 instead. On docker stop it logs SIGTERM received, draining, and exits within 10 seconds.

/health checks the configuration, and nothing else. It does not open the database, does not call a tracker, and is not rate-limited. So read a page too, with the id of a site: curl -H 'Origin: <the site>' 'http://localhost:8080/feedback?url=<the page, encoded>&client=<site_…>'.

You see	`/health`	A read	Cause
Compose starts nothing: `required variable FRUITBACK_POSTGRES_PASSWORD is missing a value`	n/a	n/a	The password is not in `.env` (measured). The same for `ALLOWED_ORIGINS`.
`docker compose up --wait` ends with `container … is unhealthy`	`503` `{"ok":false,"error":"misconfigured","missing":[…]}`	`500`	Each entry of `missing` names a variable and what to do. Fix every one: they are all listed at once (measured).
The entry names `FRUITBACK_DATABASE_URL` and « Upgrading »	`503`	`500`	No database: an `.env` of before 0.3.0, or a `docker run` with no address. See Upgrading.
An entry names a variable of before, `FRUITBACK_CLIENTS` for example	`503`	`500`	A variable an instance of before set. Its sentence says what replaces it. See Upgrading.
Nothing is planted, the widget keeps the note	`200`	`502 store-unavailable`	The database does not answer, or a tracker refuses its key. The log says which (measured with the database stopped).
No link in the log at step 4	`200`	n/a	`FRUITBACK_FIRST_ADMIN` is empty, or a workspace has an owner already. Use `admin-link`.
Pins do not appear, the browser shows a CORS error	`200`	`403 origin-not-allowed` or `origin-not-allowed-for-client`	The page is on no site of the console, or on another site than the one its script names (measured).
A read answers `400 client-required` or `400 unknown-client`	`200`	`400`	The read names no site, or a site this worker does not have. The script of the console names it (measured).
A visitor sees no pin, and a note they send is kept	`200`	`401 identity-required`	The site is read by « Members of … ». Choose « Everyone who visits » in the console, if every visitor may read the notes.
Docker says `healthy`, nothing answers from outside	unreachable	unreachable	`HOST` is not `0.0.0.0`, or `PORT` is not the published port.
Every caller gets `429` at once	`200`	`429 rate-limited`	Callers share one bucket: see Check it.

The other answers the worker gives:

Code	`error`	Means
`400`	`invalid-body`, `invalid-json`, `invalid-note`, `invalid-url`, `missing-url`	The request is wrong.
`401`	`identity-required`, `invalid-identity`	The read needs a session, or the token did not verify.
`403`	`origin-not-allowed`, `origin-not-allowed-for-client`	The calling site is not allowed.
`404`, `405`	`not-found`, `method-not-allowed`	No such route.
`413`	`payload-too-large`	The body is over the limit.
`429`	`rate-limited`	The address used up its requests for the minute.
`500`	`misconfigured`	The configuration is refused; `/health` says why.
`502`	`store-unavailable`	The store did not answer. The widget keeps the note.
`503`	`limiter-unavailable`	The rate limiter’s state did not answer.