fix discord bot config restart and startup connection retry
Container image / image (push) Failing after 1m1s

This commit is contained in:
legop3
2026-09-18 14:41:11 -04:00
parent 61ee692815
commit fe2c29371f
6 changed files with 559 additions and 283 deletions
@@ -6,12 +6,14 @@ const { MessageFlags } = require('discord.js');
function createChannelIO({ client, logger, sanitizeMentions }) { function createChannelIO({ client, logger, sanitizeMentions }) {
const channelCache = new Map(); const channelCache = new Map();
const typingMessageCache = new Map(); const typingMessageCache = new Map();
let stopped = false;
async function fetchChannel(id) { async function fetchChannel(id) {
if (!id) return null; if (stopped || !client.isReady() || !id) return null;
if (channelCache.has(id)) return channelCache.get(id); if (channelCache.has(id)) return channelCache.get(id);
try { try {
const channel = await client.channels.fetch(id); const channel = await client.channels.fetch(id);
if (stopped) return null;
if (channel) { if (channel) {
channelCache.set(id, channel); channelCache.set(id, channel);
return channel; return channel;
@@ -68,6 +70,7 @@ function createChannelIO({ client, logger, sanitizeMentions }) {
const content = `-# *${username} is typing...*`; const content = `-# *${username} is typing...*`;
try { try {
const message = await channel.send({ content, allowedMentions: { parse: [] }, flags: [MessageFlags.SuppressNotifications]}); const message = await channel.send({ content, allowedMentions: { parse: [] }, flags: [MessageFlags.SuppressNotifications]});
if (stopped) return;
const timeoutId = setTimeout(() => { const timeoutId = setTimeout(() => {
clearTypingMessage(entry.guildId, typingId); clearTypingMessage(entry.guildId, typingId);
}, 20000); }, 20000);
@@ -78,6 +81,12 @@ function createChannelIO({ client, logger, sanitizeMentions }) {
} }
return { return {
stop() {
stopped = true;
for (const record of typingMessageCache.values()) clearTimeout(record.timeoutId);
typingMessageCache.clear();
channelCache.clear();
},
fetchChannel, fetchChannel,
sendToChannel, sendToChannel,
clearTypingMessage, clearTypingMessage,
@@ -0,0 +1,94 @@
// Owns connection replacement and recovery from failed initial logins.
function createConnectionManager({ createRuntime, logger }) {
let desired = { enabled: false, token: '' };
let revision = 0;
let runtime = null;
let queue = Promise.resolve();
let retryTimer = null;
let retryDelay = 5000;
function enqueue(operation) {
const pending = queue.then(operation);
// Report failures without poisoning subsequent configuration changes.
queue = pending.catch((err) => logger.error('Discord connection update failed', err.message));
return pending;
}
async function stop() {
const previous = runtime;
runtime = null;
if (!previous) return;
previous.stop();
// discord.js cannot cancel login while gateway discovery is pending.
// Let that attempt settle before destroying it, otherwise it can open a
// socket after destroy() has already run and leave an orphan connection.
await previous.login;
await previous.client.destroy();
}
function scheduleRetry(version) {
if (version !== revision || !desired.enabled) return;
const delay = retryDelay;
retryDelay = Math.min(retryDelay * 2, 60000);
logger.warn('Retrying Discord login', { delayMs: delay });
retryTimer = setTimeout(() => {
retryTimer = null;
enqueue(() => start(version));
}, delay);
retryTimer.unref?.();
}
async function start(version) {
if (version !== revision || !desired.enabled) return;
await stop();
if (version !== revision || !desired.enabled) return;
const current = createRuntime();
runtime = current;
const { client } = current;
client.on('error', (err) => logger.error('Discord client error', err.message));
client.on('shardError', (err, shardId) => {
logger.warn('Discord connection error', { shardId, error: err.message });
});
client.on('shardDisconnect', (event, shardId) => {
logger.warn('Discord disconnected', { shardId, code: event.code });
});
client.on('clientReady', () => { retryDelay = 5000; });
// Login must not block configuration saves while the network is down.
// Each attempt has its own client, so stale login completion/cleanup cannot
// clear the credentials or stop the connection of its replacement.
current.login = client.login(desired.token).catch((err) => {
if (version !== revision || runtime !== current) return;
logger.error('Discord login failed', err.message);
enqueue(async () => {
if (version !== revision || runtime !== current) return;
await stop();
// Credentials require an operator edit; retry network/service failures.
if (err.code === 'TokenInvalid' || err.status === 401) return;
scheduleRetry(version);
});
});
}
function configure({ enabled, token }) {
const next = { enabled: Boolean(enabled), token };
if (next.enabled === desired.enabled && next.token === desired.token) return queue;
desired = next;
const version = ++revision;
clearTimeout(retryTimer);
retryTimer = null;
retryDelay = 5000;
return enqueue(async () => {
if (version !== revision) return;
if (desired.enabled) await start(version);
else await stop();
});
}
return {
configure,
refresh() { runtime?.refresh(); },
};
}
module.exports = { createConnectionManager };
@@ -0,0 +1,146 @@
const test = require('node:test');
const assert = require('node:assert/strict');
const { EventEmitter } = require('node:events');
const { createConnectionManager } = require('./connection');
const flush = () => new Promise((resolve) => setImmediate(resolve));
function deferred() {
let resolve;
let reject;
const promise = new Promise((yes, no) => { resolve = yes; reject = no; });
return { promise, resolve, reject };
}
function setup(t, attempts = []) {
t.mock.timers.enable({ apis: ['setTimeout'] });
const runtimes = [];
const logs = [];
const manager = createConnectionManager({
logger: {
warn: (...args) => logs.push(args),
error: (...args) => logs.push(args),
},
createRuntime() {
const attempt = attempts[runtimes.length] || {};
const client = new EventEmitter();
const runtime = {
client,
stopped: false,
destroyed: false,
refresh() {},
stop() { runtime.stopped = true; },
};
client.login = async (token) => {
runtime.token = token;
await attempt.login?.();
};
client.destroy = async () => {
await attempt.destroy?.();
runtime.destroyed = true;
};
runtimes.push(runtime);
return runtime;
},
});
return { manager, runtimes, logs };
}
test('retries a network failure at startup with a fresh client', async (t) => {
const { manager, runtimes } = setup(t, [{ login: () => { throw new Error('getaddrinfo EAI_AGAIN'); } }]);
await manager.configure({ enabled: true, token: 'token' });
await flush();
assert.equal(runtimes[0].stopped, true);
assert.equal(runtimes[0].destroyed, true);
t.mock.timers.tick(4999);
await flush();
assert.equal(runtimes.length, 1);
t.mock.timers.tick(1);
await flush();
assert.equal(runtimes.length, 2);
assert.equal(runtimes[1].token, 'token');
assert.notEqual(runtimes[0].client, runtimes[1].client);
});
test('backs off repeated failures and cancels retries when disabled', async (t) => {
const failed = { login: () => { throw new Error('Network unavailable'); } };
const { manager, runtimes, logs } = setup(t, Array(8).fill(failed));
await manager.configure({ enabled: true, token: 'token' });
await flush();
for (const delay of [5000, 10000, 20000, 40000, 60000]) {
t.mock.timers.tick(delay);
await flush();
}
assert.deepEqual(logs.filter(([message]) => message === 'Retrying Discord login').map(([, data]) => data.delayMs),
[5000, 10000, 20000, 40000, 60000, 60000]);
await manager.configure({ enabled: false, token: 'token' });
t.mock.timers.tick(120000);
await flush();
assert.equal(runtimes.length, 6);
});
test('waits for logout before logging in with a changed token', async (t) => {
const logout = deferred();
const { manager, runtimes } = setup(t, [{ destroy: () => logout.promise }]);
await manager.configure({ enabled: true, token: 'old' });
const change = manager.configure({ enabled: true, token: 'new' });
await flush();
assert.equal(runtimes[0].stopped, true);
assert.equal(runtimes.length, 1);
logout.resolve();
await change;
assert.equal(runtimes[0].destroyed, true);
assert.equal(runtimes[1].token, 'new');
});
test('retires a pending login before replacement and ignores its stale failure', async (t) => {
const login = deferred();
const { manager, runtimes } = setup(t, [{ login: () => login.promise }]);
await manager.configure({ enabled: true, token: 'old' });
const change = manager.configure({ enabled: true, token: 'new' });
await flush();
assert.equal(runtimes[0].stopped, true);
assert.equal(runtimes[0].destroyed, false);
login.reject(new Error('Old connection failed'));
await change;
assert.equal(runtimes[0].destroyed, true);
assert.equal(runtimes[1].token, 'new');
t.mock.timers.tick(120000);
await flush();
assert.equal(runtimes.length, 2);
});
test('re-enabling creates a fresh client; unrelated config edits do not reconnect', async (t) => {
const { manager, runtimes } = setup(t);
await manager.configure({ enabled: true, token: 'token' });
await manager.configure({ enabled: true, token: 'token', channels: { general: 'new-channel' } });
assert.equal(runtimes.length, 1);
await manager.configure({ enabled: false, token: 'token' });
assert.equal(runtimes[0].destroyed, true);
await manager.configure({ enabled: true, token: 'token' });
assert.equal(runtimes.length, 2);
});
test('invalid credentials wait for a token edit instead of retrying', async (t) => {
const { manager, runtimes } = setup(t, [{ login: () => {
throw Object.assign(new Error('Invalid token'), { code: 'TokenInvalid' });
} }]);
await manager.configure({ enabled: true, token: 'bad' });
await flush();
t.mock.timers.tick(120000);
await flush();
assert.equal(runtimes.length, 1);
await manager.configure({ enabled: true, token: 'fixed' });
assert.equal(runtimes[1].token, 'fixed');
});
test('client and gateway errors are logged without throwing', async (t) => {
const { manager, runtimes, logs } = setup(t);
await manager.configure({ enabled: true, token: 'token' });
runtimes[0].client.emit('error', new Error('Client failure'));
runtimes[0].client.emit('shardError', new Error('Gateway failure'), 0);
runtimes[0].client.emit('shardDisconnect', { code: 1006 }, 0);
assert.deepEqual(logs.map(([message]) => message), [
'Discord client error', 'Discord connection error', 'Discord disconnected',
]);
});
+38 -36
View File
@@ -67,6 +67,7 @@ const {
} = require('../privateRoverAccessRequestService'); } = require('../privateRoverAccessRequestService');
const { subscribe } = require('../eventBus'); const { subscribe } = require('../eventBus');
const { createPresenceManager } = require('./presence'); const { createPresenceManager } = require('./presence');
const { createConnectionManager } = require('./connection');
const { createChannelIO } = require('./channelIO'); const { createChannelIO } = require('./channelIO');
const { createCommandHandlers } = require('../operatorCommandService'); const { createCommandHandlers } = require('../operatorCommandService');
const greenModeService = require('../greenModeService'); const greenModeService = require('../greenModeService');
@@ -86,9 +87,7 @@ const {
buildStatusMessage, buildStatusMessage,
} = require('../replayDeliveryService/workflow'); } = require('../replayDeliveryService/workflow');
// Discord helper modules retain references to these objects. Mutating those // Keep configuration object identities stable for the active runtime's helpers.
// references on configuration application updates command and integration
// behavior without registering a second tree of Discord/event listeners.
const config = structuredClone(loadConfig()); const config = structuredClone(loadConfig());
const discordConfig = config.discord || {}; const discordConfig = config.discord || {};
let enabled = Boolean(discordConfig.enabled); let enabled = Boolean(discordConfig.enabled);
@@ -106,11 +105,6 @@ const intents = [
GatewayIntentBits.MessageContent, GatewayIntentBits.MessageContent,
]; ];
const client = new Client({
intents,
partials: [Partials.Channel, Partials.Message, Partials.Reaction, Partials.User],
});
function sanitizeMentions(text) { function sanitizeMentions(text) {
if (!text) return ''; if (!text) return '';
return String(text) return String(text)
@@ -156,6 +150,13 @@ function countReady() {
return { ready, total }; return { ready, total };
} }
// Each connection owns its listeners, timers, and cached Discord objects.
// Rebuilding them prevents a reconnect from retaining a destroyed client.
function createDiscordRuntime() {
const client = new Client({
intents,
partials: [Partials.Channel, Partials.Message, Partials.Reaction, Partials.User],
});
const channelIO = createChannelIO({ const channelIO = createChannelIO({
client, client,
logger, logger,
@@ -181,7 +182,7 @@ const replayCaption = createReplayCaptionBuilder({
// Discord is the preferred replay host only while this optional feature is // Discord is the preferred replay host only while this optional feature is
// active. The core replay delivery service owns generation and automatically // active. The core replay delivery service owns generation and automatically
// falls back to its local media store when any operation below fails. // falls back to its local media store when any operation below fails.
registerPreferredDeliveryProvider({ const unregisterReplayProvider = registerPreferredDeliveryProvider({
async begin(job) { async begin(job) {
const channelId = discordConfig.channels?.replay; const channelId = discordConfig.channels?.replay;
if (!enabled || !channelId) throw new Error('Discord replay delivery is disabled'); if (!enabled || !channelId) throw new Error('Discord replay delivery is disabled');
@@ -405,7 +406,7 @@ function restartFleetDailyReports() {
fleetDailyReports.start(); fleetDailyReports.start();
} }
client.on('ready', () => { client.on('clientReady', () => {
logger.info('Discord bot logged in', { tag: client.user?.tag }); logger.info('Discord bot logged in', { tag: client.user?.tag });
presence.schedulePresenceRotation(); presence.schedulePresenceRotation();
// Discord is only a delivery consumer. Starting its scheduler after the bot // Discord is only a delivery consumer. Starting its scheduler after the bot
@@ -414,42 +415,45 @@ client.on('ready', () => {
restartFleetDailyReports(); restartFleetDailyReports();
}); });
return {
client,
refresh() {
if (client.isReady()) restartFleetDailyReports();
},
stop() {
for (const event of ['clientReady', 'messageCreate', 'typingStart', 'messageReactionAdd']) {
client.removeAllListeners(event);
}
integrationHandlers.stop();
unregisterReplayProvider();
presence.stop();
channelIO.stop();
fleetDailyReports?.stop();
},
};
}
const connection = createConnectionManager({ createRuntime: createDiscordRuntime, logger });
function replaceObject(target, source = {}) { function replaceObject(target, source = {}) {
Object.keys(target).forEach((key) => delete target[key]); Object.keys(target).forEach((key) => delete target[key]);
Object.assign(target, structuredClone(source)); Object.assign(target, structuredClone(source));
} }
function applyDiscordConfig(nextDiscordConfig = {}) { async function applyDiscordConfig(nextDiscordConfig = {}) {
const wasEnabled = enabled;
const previousToken = discordConfig.token;
replaceObject(discordConfig, nextDiscordConfig); replaceObject(discordConfig, nextDiscordConfig);
config.discord = discordConfig; config.discord = discordConfig;
enabled = Boolean(discordConfig.enabled); enabled = Boolean(discordConfig.enabled);
await connection.configure(discordConfig);
if (!enabled) { connection.refresh();
fleetDailyReports?.stop();
fleetDailyReports = null;
if (wasEnabled) client.destroy();
return;
}
if (!wasEnabled || previousToken !== discordConfig.token) {
if (wasEnabled) client.destroy();
// Login health is reported by Discord itself; do not hold the committed
// configuration request open while an external network service connects.
client.login(discordConfig.token).catch((err) => {
logger.error('Discord login failed after configuration change', err.message);
});
} else if (client.isReady()) {
restartFleetDailyReports();
}
} }
function applySharedConfigSection(section, value) { function applySharedConfigSection(section, value) {
config[section] = structuredClone(value); config[section] = structuredClone(value);
// Fleet delivery owns a timer derived from both Discord and fleet settings. // Fleet delivery owns a timer derived from both Discord and fleet settings.
// Reconnecting is unnecessary; rebuild only that scheduler when ready. // Reconnecting is unnecessary; rebuild only that scheduler when ready.
if (section === 'fleetReports' && client.isReady()) { if (section === 'fleetReports') {
restartFleetDailyReports(); connection.refresh();
} }
} }
@@ -458,10 +462,8 @@ registerConfigurationHandler('discord', applyDiscordConfig);
registerConfigurationHandler(section, (value) => applySharedConfigSection(section, value)); registerConfigurationHandler(section, (value) => applySharedConfigSection(section, value));
}); });
if (enabled) { connection.configure(discordConfig).catch((err) => {
client.login(discordConfig.token).catch((err) => { logger.error('Discord startup failed', err.message);
logger.error('Discord login failed', err.message);
}); });
}
module.exports = {}; module.exports = {};
@@ -24,6 +24,7 @@ function createIntegrations(deps) {
const userAnnouncements = createUserAnnouncements({ ...deps, sendToChannel, schedulePresenceRotation }); const userAnnouncements = createUserAnnouncements({ ...deps, sendToChannel, schedulePresenceRotation });
function register() { function register() {
let stopped = false;
client.on('typingStart', (typing) => { client.on('typingStart', (typing) => {
chat.handleDiscordTypingStart(typing).catch((err) => logger.warn('Error handling Discord typing', err.message)); chat.handleDiscordTypingStart(typing).catch((err) => logger.warn('Error handling Discord typing', err.message));
}); });
@@ -33,14 +34,31 @@ function createIntegrations(deps) {
dm.handlePrivateAccessReaction(reaction, user).catch((err) => logger.warn('Error handling private access reaction', err.message)); dm.handlePrivateAccessReaction(reaction, user).catch((err) => logger.warn('Error handling private access reaction', err.message));
}); });
subscribe('*', handleBusEvent); // Only the active, ready connection consumes server events. Retain each
subscribe('*', userAnnouncements.handleBusEvent); // unsubscribe function so replacement clients never duplicate deliveries.
subscribe('verification.requested', dm.sendVerificationRequestDms); const unsubscribe = [];
subscribe('privateRoverAccess.requested', dm.sendPrivateRoverAccessRequestDms); function subscribeWhileReady(type, handler) {
subscribe('chat:message', chat.handleChatBridgeOutbound); unsubscribe.push(subscribe(type, (event) => {
subscribe('chat:typing', chat.handleChatTypingOutbound); Promise.resolve().then(() => {
if (!stopped && client.isReady()) return handler(event);
}).catch((err) => {
logger.warn('Error handling Discord integration event', { type, error: err.message });
});
}));
}
subscribeWhileReady('*', handleBusEvent);
subscribeWhileReady('*', userAnnouncements.handleBusEvent);
subscribeWhileReady('verification.requested', dm.sendVerificationRequestDms);
subscribeWhileReady('privateRoverAccess.requested', dm.sendPrivateRoverAccessRequestDms);
subscribeWhileReady('chat:message', chat.handleChatBridgeOutbound);
subscribeWhileReady('chat:typing', chat.handleChatTypingOutbound);
return { return {
stop() {
stopped = true;
unsubscribe.forEach((remove) => remove());
},
handleBridgeInbound: chat.handleBridgeInbound, handleBridgeInbound: chat.handleBridgeInbound,
handleBusEvent, handleBusEvent,
}; };
@@ -7,6 +7,7 @@ function createPresenceManager({ client, logger, getMode, getGlobalObjective, co
const PRESENCE_ROTATE_MS = 20000; const PRESENCE_ROTATE_MS = 20000;
let presenceInterval = null; let presenceInterval = null;
let presenceShowObjective = false; let presenceShowObjective = false;
let stopped = false;
function truncatePresenceText(text, maxLength) { function truncatePresenceText(text, maxLength) {
if (!text) return ''; if (!text) return '';
@@ -40,6 +41,7 @@ function createPresenceManager({ client, logger, getMode, getGlobalObjective, co
} }
function schedulePresenceRotation() { function schedulePresenceRotation() {
if (stopped) return;
if (presenceInterval) { if (presenceInterval) {
clearInterval(presenceInterval); clearInterval(presenceInterval);
presenceInterval = null; presenceInterval = null;
@@ -59,6 +61,11 @@ function createPresenceManager({ client, logger, getMode, getGlobalObjective, co
} }
return { return {
stop() {
stopped = true;
clearInterval(presenceInterval);
presenceInterval = null;
},
schedulePresenceRotation, schedulePresenceRotation,
}; };
} }