screentinker/server/services/heartbeat.js
ScreenTinker 8ad2258e7c feat: app-ending signal (exit-signal contract v1) — server + APK + .wgt + /player
Best-effort "last gasp" so Offline is annotated with WHY it went away — completing the liveness story.
Categories: crashed (client uncaught-exception), clean_exit (client confident lifecycle-end, best-effort),
silent (SERVER-inferred by absence — the honest catch-all for violent/external death incl. force-stop/MDM).

SERVER:
- device:exit socket handler + token-authed beacon POST /api/device/exit (reliable-on-unload). Both gated
  by liveness.sanitizeExitReason (honesty: only crashed/clean_exit accepted; 'silent'/unknown rejected).
- offline_reason/offline_reason_at/offline_detail columns (additive migration). Clear-on-online (a reason
  is always THIS session's); offline transition COALESCEs to 'silent'. Pure annotation — offline detection
  and #148/liveness are untouched. Offline dashboard emits carry offline_reason + client_type.
CLIENTS (canonical {reason,detail} shape):
- /player: window error/unhandledrejection + pagehide(persisted=false) -> sendBeacon.
- .wgt: same + BACK-key exit -> socket.emit + sendBeacon.
- APK: global UncaughtExceptionHandler -> crashed (blocking beacon, chains to default); Service.onDestroy
  -> clean_exit (socket + bounded beacon). New ExitSignal.kt. onStop/onPause NOT wired (background != exit).
Proven (Phase 3): per-category classification, nothing misclassified, external kill -> silent (never
clean_exit), backgrounding emits no false exit, #148/reconnect-vs-exit intact. 382/382 suite green.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-08 15:32:40 -05:00

250 lines
13 KiB
JavaScript

const { db, pruneStatusLog } = require('../db/database');
const config = require('../config');
const { deviceRoom, emitToWorkspace } = require('../lib/socket-rooms');
const statusLogWriter = require('../lib/status-log-writer');
const { chunkedDelete, currentBand, yieldTick } = require('../lib/chunked-prune'); // #146 non-blocking sweeps
const liveness = require('../lib/liveness'); // v4 core pass: server-derived 3-state liveness
// Track connected device sockets: deviceId -> { socketId, lastHeartbeat }
const deviceConnections = new Map();
// FIX 2: version-agnostic reconnect-frequency signal (every client reconnects the same way). A
// rolling window of recent (re)register timestamps per device -> "degraded-reconnecting" when it churns.
let _io = null; // captured in startHeartbeatChecker so livenessFor() can check namespace presence
const RECONNECT_WINDOW_MS = 60000;
const reconnectTimes = new Map(); // deviceId -> [timestamps within the window]
function recordReconnect(deviceId, now = Date.now()) {
const arr = (reconnectTimes.get(deviceId) || []).filter(t => now - t < RECONNECT_WINDOW_MS);
arr.push(now);
reconnectTimes.set(deviceId, arr);
}
function recentReconnects(deviceId, now = Date.now()) {
const arr = (reconnectTimes.get(deviceId) || []).filter(t => now - t < RECONNECT_WINDOW_MS);
if (arr.length) reconnectTimes.set(deviceId, arr); else reconnectTimes.delete(deviceId);
return arr.length;
}
// Server-derived liveness for a device — from socket presence + heartbeat age + reconnect churn ONLY
// (all version-agnostic). A disconnected device is a clean 'offline' (normal state, not an error).
function livenessFor(deviceId) {
const conn = deviceConnections.get(deviceId);
const deviceNs = _io ? _io.of('/device') : null;
const connected = !!(conn && deviceNs && deviceNs.sockets.has(conn.socketId));
const lastHeartbeatAgeMs = conn ? (Date.now() - conn.lastHeartbeat) : Infinity;
return liveness.deriveLiveness({ connected, lastHeartbeatAgeMs, recentReconnects: recentReconnects(deviceId) });
}
function startHeartbeatChecker(io) {
_io = io; // FIX 2: for livenessFor() namespace-presence checks
// #146: startup sweep is chunked + async + fire-and-forget + NOT band-gated, so a
// bloated device_status_log self-heals on next deploy WITHOUT freezing boot (the old
// whole-table sort froze boot 40-48s -> healthcheck fail -> restart loop). It
// trickles in bounded batches while the server comes up and serves.
pruneStatusLog({ bandGate: false }).catch(() => {});
// #146: start the batched device_status_log flush loop.
statusLogWriter.start();
const deviceNs = io.of('/device');
setInterval(() => {
const now = Date.now();
const dashboardNs = io.of('/dashboard');
// #146 BILLING: credit currently-connected devices' usage for this interval.
// Fire-and-forget + never throws into the interval (billing must not perturb the
// heartbeat). Reads the same live presence map as the offline check below.
accrueUsage(now).catch(() => {});
// Check database for devices that should be offline
const onlineDevices = db.prepare("SELECT id, last_heartbeat FROM devices WHERE status = 'online'").all();
for (const device of onlineDevices) {
const conn = deviceConnections.get(device.id);
// #146: a device with a live, still-connected socket is UP, even if its last
// heartbeat event is stuck behind a lagged event loop. Marking it offline on a
// stale in-memory lastHeartbeat was the second false-offline cause (the screen
// is online and playing, the CMS says offline). The socket still being in the
// /device namespace is the authoritative liveness signal — trust it over the
// (possibly queued) heartbeat clock. If the socket is genuinely gone, conn is
// either absent or points at a socket no longer in the namespace, and we fall
// through to the timeout below.
if (conn && deviceNs.sockets.has(conn.socketId)) continue;
const lastBeat = conn ? conn.lastHeartbeat : (device.last_heartbeat ? device.last_heartbeat * 1000 : 0);
if (now - lastBeat > config.heartbeatTimeout) {
// #148 Item 2: marking a device offline MUST also close any socket we still hold for
// it, so DB-offline can never diverge from socket-state into a silent half-open the
// client is never told about. The live-socket guard above already `continue`d for a
// genuinely-live socket, so this only reaps a stale/half-open one (Engine.IO's
// ping-timeout also reaps it, but this makes offline<=>closed explicit + immediate).
if (conn) {
const sock = deviceNs.sockets.get(conn.socketId);
if (sock) { try { sock.disconnect(true); } catch (_) { /* already gone */ } }
}
// Exit-signal contract: this timeout path is the classic 'silent' case (froze, no clean
// disconnect, no signal) — COALESCE annotates 'silent' unless a device:exit reason arrived
// this session (e.g. a crash emit that beat the freeze). Pure annotation; detection unchanged.
db.prepare("UPDATE devices SET status = 'offline', updated_at = strftime('%s','now'), offline_reason = COALESCE(offline_reason, 'silent'), offline_reason_at = COALESCE(offline_reason_at, strftime('%s','now')) WHERE id = ?")
.run(device.id);
deviceConnections.delete(device.id);
const _off = db.prepare("SELECT offline_reason, offline_detail, client_type FROM devices WHERE id = ?").get(device.id) || {};
// Notify dashboard (workspace-scoped via the device's room).
emitToWorkspace(dashboardNs, deviceRoom(device.id), 'dashboard:device-status', {
device_id: device.id,
status: 'offline',
liveness: 'offline', // FIX 2: derived — no live socket => offline (a normal state, not an error)
offline_reason: _off.offline_reason || 'silent', // exit-signal contract: manner-of-death
offline_detail: _off.offline_detail || null,
client_type: _off.client_type || null,
telemetry: null
});
reconnectTimes.delete(device.id); // clear churn history on a clean offline
console.log(`Device ${device.id} marked offline (heartbeat timeout)`);
// #146: batch through the coalescing writer (was an immediate INSERT here).
statusLogWriter.record(device.id, 'offline_timeout');
}
}
// #146: all table-growth maintenance runs OFF the interval body — async, chunked,
// band-gated, re-entrancy-guarded — so a sweep can never block the loop or stack.
// The offline-marking above stays synchronous (it's the core heartbeat function).
runMaintenance();
}, config.heartbeatInterval);
}
// #146: batched play-log prune (idx_play_logs_time), chunked so a 90-day backlog
// trims across many bounded DELETEs instead of one large statement.
const _delPlayLogs = db.prepare('DELETE FROM play_logs WHERE rowid IN (SELECT rowid FROM play_logs WHERE started_at < ? LIMIT ?)');
async function prunePlayLogs() {
const cutoff = Math.floor(Date.now() / 1000) - (90 * 86400);
return (await chunkedDelete((lim) => _delPlayLogs.run(cutoff, lim).changes, { batch: config.statusLogPruneBatch })).deleted;
}
// #146 interval maintenance — band-gated (skip while loaded; runs next tick) and
// re-entrancy-guarded (a long run never stacks with the next interval). Never throws
// into the interval. NOT for startup (see the un-gated startup prune above).
let _maintRunning = false;
async function runMaintenance() {
if (_maintRunning) return;
if (config.maintenanceBandGateEnabled && currentBand() !== 'normal') return; // #146 P1.3 kill switch
_maintRunning = true;
try {
await pruneProvisioningDevices();
await prunePlayLogs();
await pruneStatusLog({ bandGate: true }); // per-device chunked; own re-entrancy
await pruneUsageDaily(); // #146 BILLING rollup retention (chunked)
// Expiry sweeps on small tables — single cheap statements, bounded by table size.
db.prepare("DELETE FROM team_invites WHERE expires_at < strftime('%s','now')").run();
db.prepare("DELETE FROM workspace_invites WHERE expires_at < strftime('%s','now')").run();
} catch (_) { /* maintenance must never crash the interval */ } finally { _maintRunning = false; }
}
function registerConnection(deviceId, socketId) {
deviceConnections.set(deviceId, { socketId, lastHeartbeat: Date.now() });
}
function updateHeartbeat(deviceId) {
const conn = deviceConnections.get(deviceId);
if (conn) conn.lastHeartbeat = Date.now();
}
function removeConnection(deviceId) {
deviceConnections.delete(deviceId);
}
function getConnection(deviceId) {
return deviceConnections.get(deviceId);
}
function getAllConnections() {
return deviceConnections;
}
// #146: LIVE connected-device count — the set with a live socket THIS INSTANT. Cheap
// in-memory read. Distinct from devices.status='online' (persisted, lags by the
// offline-timeout). Surfaced as /api/status.devices_connected.
function getConnectedCount() {
return deviceConnections.size;
}
// #146 BILLING accumulator — credit each currently-connected device's today-row with the
// seconds elapsed since the last accrual. Retention-INDEPENDENT: it reuses the SAME live
// presence map as devices_connected (never reconstructs online time from status_log,
// which is only 3-day). Cheap + non-blocking: chunked UPSERTs, one bounded transaction
// per chunk, yielding between chunks. The per-accrual credit is CAPPED (accrualCapSeconds)
// so a stalled loop or restart gap can't inject a bogus large credit; the DAILY total is
// capped at 86400 in the UPSERT itself. Day is the UTC calendar day of the tick.
const _usageUpsert = db.prepare(`
INSERT INTO device_usage_daily (device_id, day, online_seconds) VALUES (?, ?, ?)
ON CONFLICT(device_id, day) DO UPDATE SET online_seconds = MIN(86400, online_seconds + excluded.online_seconds)
`);
let _lastAccrue = 0;
let _accrualRunning = false;
async function accrueUsage(now = Date.now()) {
if (_accrualRunning) return 0; // never stack; elapsed-based credit self-heals a skipped tick
if (_lastAccrue === 0) { _lastAccrue = now; return 0; } // first tick establishes the baseline; credit nothing
const credit = Math.min(Math.floor((now - _lastAccrue) / 1000), config.billing.accrualCapSeconds);
_lastAccrue = now;
if (credit <= 0) return 0;
const ids = Array.from(deviceConnections.keys());
if (!ids.length) return 0;
const day = new Date(now).toISOString().slice(0, 10);
_accrualRunning = true;
try {
const upsertMany = db.transaction((slice) => { for (const id of slice) _usageUpsert.run(id, day, credit); });
const batch = config.billing.accrualBatch;
for (let i = 0; i < ids.length; i += batch) {
upsertMany(ids.slice(i, i + batch));
if (i + batch < ids.length) await yieldTick(); // keep a huge fleet's accrual off the event loop
}
} finally { _accrualRunning = false; }
return ids.length;
}
// #146 BILLING: prune the daily rollup beyond retention (chunked, so it can never
// bloat-then-freeze). `day` is a sortable 'YYYY-MM-DD' string → lexical < is a date <.
const _delUsage = db.prepare('DELETE FROM device_usage_daily WHERE rowid IN (SELECT rowid FROM device_usage_daily WHERE day < ? LIMIT ?)');
async function pruneUsageDaily() {
const cutoff = new Date(Date.now() - config.billing.usageRetentionDays * 86400 * 1000).toISOString().slice(0, 10);
return (await chunkedDelete((lim) => _delUsage.run(cutoff, lim).changes, { batch: config.statusLogPruneBatch })).deleted;
}
// #142: sweep unclaimed provisioning devices older than 24h (imported devices keep a
// user_id and are preserved). #146: now async + CHUNKED (rides idx_devices_provisioning)
// so a provisioning-junk flood can't delete-cascade a huge batch in one synchronous
// statement. Returns rows deleted. NOTE: async now — callers must await.
const _delProvisioning = db.prepare(`
DELETE FROM devices WHERE rowid IN (
SELECT rowid FROM devices
WHERE status = 'provisioning' AND user_id IS NULL AND created_at < ?
LIMIT ?
)
`);
async function pruneProvisioningDevices() {
const cutoff = Math.floor(Date.now() / 1000) - (24 * 3600);
return (await chunkedDelete((lim) => _delProvisioning.run(cutoff, lim).changes, { batch: config.statusLogPruneBatch })).deleted;
}
module.exports = {
startHeartbeatChecker,
registerConnection,
updateHeartbeat,
removeConnection,
getConnection,
getAllConnections,
getConnectedCount,
recordReconnect, // FIX 2
recentReconnects, // FIX 2
livenessFor, // FIX 2
pruneProvisioningDevices,
accrueUsage,
pruneUsageDaily,
__resetAccrual: () => { _lastAccrue = 0; }, // #146 test hook: reset the accrual baseline
};