The module half of the read path. Seven tables, an ingest cursor, four public routes, and one file whose only job is deciding who may see what. **The record and the window are different things.** `rust_player_wipe_stats` and `rust_gather_totals` are permanent and per-wipe, so all-time is those rows SUMmed rather than a second set of counters that can disagree with them — that is R12's "per-wipe detail plus all-time rollups" in one table instead of two. `rust_events` is a bounded 30-day window of raw frames for the killfeed, and `rust_presence` is a board: replaced wholesale, never appended. **The feed is a cursor, not a socket, and the header says why.** Core runs Node 20, where a global WebSocket is still behind a flag, so a socket means taking `ws` — against a release that asserts it has no runtime dependencies (D5). The deciding argument is the other one though: a socket needs a cursor anyway, for whatever it missed while the module was restarting, and the catch-up path is the one that has to be right. A cursor alone is one mechanism exercised every five seconds rather than two where the second only runs after an outage. **The cursor advances after the batch, never before.** A crash between the two re-reads events already counted, which inflates a total; the other order loses them silently and for ever. One is visible and bounded, the other is invisible and permanent, so the code fails in the visible direction. A server with no cursor starts at the sidecar's current END rather than at zero — replaying a fortnight of deaths into stats for wipes the site never saw is not a catch-up. **`catalogue.js` is a security boundary, default-deny.** Protocol 2 carries IP addresses (login attempts, approvals, bans), one player's report about another, and the grid reference of somebody's base. They are stored, because an operator chasing ban evasion needs them; they are not served below the admin tier. The allowlist lives here rather than as a field on the wire, because a boundary declared by the sender is one a compromised or merely out-of-date game host can widen — the same reason core's own shard fan-out filters on the serving side. A kind this build has never heard of is not public, and a test holds the list against PROTOCOL.md §8.4 so that adding a kind to the protocol without classifying it fails a build. `PROTOCOL_VERSION` goes to 2 here in the same change as the emitters, though this module consumes none of the new frames yet: the sidecar refuses a mismatched client with a 409, so a module left on 1 would stop being able to read the board it has been reading all along. A constant that lags the deployment is an outage with a version number on it. 95 server tests, 20 client tests, every guard green, and `routes.manifest.json` regenerated against a real core at the pinned ref: 10 routes, all documented, none of core's moved. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016wDDVXWMDz82WqE1i969r4
204 lines
8.5 KiB
JavaScript
204 lines
8.5 KiB
JavaScript
// ── The near end of a call whose far end is a Rust server ─────────────────
|
|
//
|
|
// Every other file in this module reads its own tables. This one is different in
|
|
// kind: it is the only place that leaves the process.
|
|
//
|
|
// **The website process never opens a connection to a game server**
|
|
// (MODULE_API.md §2.7). It opens one to a `rust-link` sidecar, which owns the
|
|
// socket to the game, persists what the game says before forwarding it, and
|
|
// answers reads from that store. `test/noGameConnection.test.js` enforces the
|
|
// decidable half of that rule and names this file as the one that may reach the
|
|
// network:
|
|
//
|
|
// const MAY_OPEN_SOCKETS = new Set(['sidecarClient.js'])
|
|
//
|
|
// ── One client per configured server ──────────────────────────────────────
|
|
//
|
|
// R8: the bridge is one game server to one sidecar. So this file takes the
|
|
// server row as an argument rather than holding a single configured endpoint —
|
|
// six servers is six base URLs and six tokens, and core never learns there is
|
|
// more than one.
|
|
//
|
|
// ── TIMEOUT_MS is not a tuning knob. It is half of a rule. ────────────────
|
|
//
|
|
// An event action declares `budgetMs`, and core's dispatcher enforces it: when
|
|
// the budget expires it stops waiting and classifies the failure as **retry**,
|
|
// unconditionally, without asking the action — it cannot ask, the action is still
|
|
// awaiting a socket. So if core's deadline is shorter than this one, an action
|
|
// never gets to classify its own failure and `{ ok: false, retry: false }` is
|
|
// unreachable code. `budgetMs` must EXCEED this.
|
|
//
|
|
// It is also bounded from the other side: the sidecar's own RPC reply timeout is
|
|
// ten seconds, so a value below that would give up while the sidecar is still
|
|
// legitimately waiting for the game. The ordering is
|
|
// `sidecar RPC timeout < TIMEOUT_MS < budgetMs`, and every one of the three
|
|
// is written down somewhere the other two can be checked against.
|
|
//
|
|
// ── This file never throws ────────────────────────────────────────────────
|
|
//
|
|
// Every call answers `{ ok, status, data }`. A module that let a socket failure
|
|
// escape into a controller would hand an exception to a page whose whole job is
|
|
// to render while the game is off. The public site degrades; it does not 500.
|
|
|
|
const core = require('./core')
|
|
|
|
const log = core.logger('sidecar')
|
|
|
|
/** How long this client waits before giving up on a sidecar. See the header. */
|
|
const TIMEOUT_MS = 12000
|
|
|
|
/**
|
|
* The wire version this module speaks. Declared in FOUR places that must agree:
|
|
* here, `PROTOCOL_VERSION` in the sidecar, `ProtocolVersion` in the bridge
|
|
* plugin, and `protocol` in its `overlay.toml`.
|
|
*
|
|
* **2 — the read path.** The bump lands here in the same change as the emitters,
|
|
* even though this module does not yet consume any of the new frames: the
|
|
* sidecar refuses a client declaring a different version with a `409`, so a
|
|
* module left on 1 would stop being able to read the server board it has been
|
|
* reading all along. A constant that lags the deployment is not a safe default;
|
|
* it is an outage with a version number on it.
|
|
*
|
|
* It is sent on every request as `X-RustLink-Version`, which turns a mismatched
|
|
* deployment into a `409` naming both numbers instead of a parse failure three
|
|
* layers further in.
|
|
*/
|
|
const PROTOCOL_VERSION = 2
|
|
|
|
/** What a caller gets back. Shaped once so every call site reads the same. */
|
|
function reply(ok, status, data = null) {
|
|
return { ok, status, data }
|
|
}
|
|
|
|
/**
|
|
* Normalises a configured base URL into something `new URL(path, base)` will not
|
|
* surprise anybody with.
|
|
*
|
|
* A trailing slash on the base and a leading slash on the path is the classic
|
|
* way to lose a path segment, and an operator pasting a URL out of a terminal
|
|
* supplies the trailing slash about half the time.
|
|
*/
|
|
function joinUrl(baseUrl, path) {
|
|
return `${String(baseUrl).replace(/\/+$/, '')}${path}`
|
|
}
|
|
|
|
/**
|
|
* One request to one sidecar.
|
|
*
|
|
* @param {object} server a `rust_servers` row, token already decrypted
|
|
* @param {string} server.baseUrl
|
|
* @param {string|null} server.token
|
|
* @param {string} path e.g. `/server`
|
|
* @param {object} [options]
|
|
* @param {string} [options.method]
|
|
* @param {object} [options.body]
|
|
*/
|
|
async function request(server, path, { method = 'GET', body = null } = {}) {
|
|
if (!server || !server.baseUrl) return reply(false, 'not-configured')
|
|
|
|
// A sidecar with auth off does not exist — it generates and persists a token on
|
|
// first start — so a missing token here is a half-finished admin form, not a
|
|
// sidecar to try unauthenticated. Saying so beats a 401 the operator has to
|
|
// interpret.
|
|
if (!server.token) return reply(false, 'no-token')
|
|
|
|
const controller = new AbortController()
|
|
const timer = setTimeout(() => controller.abort(), TIMEOUT_MS)
|
|
|
|
try {
|
|
const res = await fetch(joinUrl(server.baseUrl, path), {
|
|
method,
|
|
signal: controller.signal,
|
|
headers: {
|
|
Authorization: `Bearer ${server.token}`,
|
|
'X-RustLink-Version': String(PROTOCOL_VERSION),
|
|
...(body ? { 'Content-Type': 'application/json' } : {}),
|
|
},
|
|
...(body ? { body: JSON.stringify(body) } : {}),
|
|
})
|
|
|
|
// A protocol mismatch is a deployment fault and deserves its own status, not
|
|
// to be folded into "the sidecar said no". The operator's fix is an upgrade
|
|
// of one component, and the message has to be able to say which.
|
|
if (res.status === 409) {
|
|
const detail = await safeJson(res)
|
|
log.warn('protocol mismatch', {
|
|
server: server.id,
|
|
module: PROTOCOL_VERSION,
|
|
sidecar: detail && detail.sidecar_protocol,
|
|
})
|
|
return reply(false, 'protocol-mismatch', detail)
|
|
}
|
|
|
|
if (res.status === 401) return reply(false, 'unauthorized')
|
|
|
|
// 204 is an ANSWER, not an absence of one: the sidecar is up and reports that
|
|
// the game has never connected. Collapsing it into a failure would make a
|
|
// freshly installed server indistinguishable from an unreachable one.
|
|
if (res.status === 204) return reply(true, 'empty', null)
|
|
|
|
if (!res.ok) return reply(false, `http-${res.status}`)
|
|
|
|
return reply(true, 'ok', await safeJson(res))
|
|
} catch (err) {
|
|
// `AbortError` is this client's own deadline firing, and it is worth telling
|
|
// apart from a refused connection: one means the sidecar is slow or the game
|
|
// is not answering, the other means nothing is listening.
|
|
const status = err && err.name === 'AbortError' ? 'timeout' : 'transport-error'
|
|
log.warn('sidecar request failed', { server: server.id, path, status, error: err.message })
|
|
return reply(false, status)
|
|
} finally {
|
|
clearTimeout(timer)
|
|
}
|
|
}
|
|
|
|
async function safeJson(res) {
|
|
try {
|
|
return await res.json()
|
|
} catch {
|
|
// A sidecar that answered 200 with something that is not JSON is a sidecar
|
|
// this module cannot use, but it is not a reason to throw at a page.
|
|
return null
|
|
}
|
|
}
|
|
|
|
/** Liveness, the protocol version, and whether the plugin is connected. Unauthenticated at the far end, but sent authenticated anyway so one code path covers every call. */
|
|
const health = (server) => request(server, '/health')
|
|
|
|
/** The last `server.hello` the sidecar stored. Answers while the game is off. */
|
|
const serverBoard = (server) => request(server, '/server')
|
|
|
|
/** A live round trip through the sidecar to the game. Fails when the game is down, by design. */
|
|
const liveStatus = (server) => request(server, '/status')
|
|
|
|
/** Every board at once: what is true now, before following what happens next. */
|
|
const boards = (server) => request(server, '/boards')
|
|
|
|
/**
|
|
* The ingest cursor: events after `since`, oldest first.
|
|
*
|
|
* **`since` is required here, unlike on the wire.** The sidecar treats an omitted
|
|
* cursor as "tell me where the end is", which is a genuinely useful question and
|
|
* a catastrophic default for an ingest loop that would silently store nothing
|
|
* and advance past everything. So the question is asked explicitly, by name, and
|
|
* a caller cannot get it by forgetting an argument.
|
|
*/
|
|
const feed = (server, since, limit = 200) =>
|
|
request(server, `/feed?since=${encodeURIComponent(since)}&limit=${encodeURIComponent(limit)}`)
|
|
|
|
/** Where the sidecar's history currently ends. What a new server's cursor starts at. */
|
|
const feedTail = (server) => request(server, '/feed')
|
|
|
|
module.exports = {
|
|
TIMEOUT_MS,
|
|
PROTOCOL_VERSION,
|
|
request,
|
|
health,
|
|
serverBoard,
|
|
liveStatus,
|
|
boards,
|
|
feed,
|
|
feedTail,
|
|
joinUrl,
|
|
}
|