cfsa-antigravity 1.0.0

package/template/.agent/skill-library/stack/queue/bullmq/SKILL.md

@@ -0,0 +1,385 @@
+---
+name: bullmq
+description: Redis-based job queue with BullMQ. Use when building background job processing, task scheduling, rate-limited workflows, or distributed job pipelines with Node.js. Triggers on BullMQ, Queue, Worker, job queue, background jobs, delayed jobs, repeatable jobs, Bull Board.
+version: 1.0.0
+---
+
+# BullMQ
+
+BullMQ is a Redis-based queue for Node.js. It provides reliable job processing with retries, delays, priorities, rate limiting, repeatable jobs, parent-child flows, and sandboxed processors. It is the successor to Bull and the standard choice for production job queues in the Node.js ecosystem.
+
+## Installation
+
+```bash
+pnpm add bullmq
+# Redis is required — BullMQ connects to Redis, not an embedded store
+pnpm add ioredis  # peer dependency
+```
+
+## Core Architecture
+
+```
+Producer (Queue) ──enqueue──> Redis <──dequeue── Consumer (Worker)
+                                │
+                         QueueEvents (listener)
+```
+
+Three primary classes:
+- **Queue**: Enqueues jobs (producer side)
+- **Worker**: Processes jobs (consumer side)
+- **QueueEvents**: Listens to lifecycle events (monitoring side)
+
+## Connection Configuration
+
+```typescript
+import { Queue, Worker } from 'bullmq';
+import IORedis from 'ioredis';
+
+// Shared connection — reuse across Queue, Worker, QueueEvents
+const connection = new IORedis({
+  host: process.env.REDIS_HOST ?? '127.0.0.1',
+  port: Number(process.env.REDIS_PORT ?? 6379),
+  password: process.env.REDIS_PASSWORD,
+  maxRetriesPerRequest: null, // Required for BullMQ workers
+});
+
+const queue = new Queue('email', { connection });
+const worker = new Worker('email', processEmail, { connection });
+```
+
+**Anti-pattern**: Never set `maxRetriesPerRequest` to a number on worker connections. BullMQ uses blocking Redis commands (`BRPOPLPUSH`) that must not be retried automatically.
+
+## Queue — Producing Jobs
+
+```typescript
+import { Queue } from 'bullmq';
+
+const emailQueue = new Queue('email', { connection });
+
+// Basic job
+await emailQueue.add('welcome', { userId: 'u_123', template: 'welcome' });
+
+// Job with options
+await emailQueue.add('invoice', { orderId: 'ord_456' }, {
+  delay: 5000,              // Wait 5s before processing
+  attempts: 3,              // Retry up to 3 times on failure
+  backoff: {
+    type: 'exponential',    // 'fixed' | 'exponential'
+    delay: 1000,            // Base delay in ms
+  },
+  priority: 1,              // Lower number = higher priority (default: 0)
+  removeOnComplete: { count: 1000 },  // Keep last 1000 completed
+  removeOnFail: { age: 7 * 24 * 3600 }, // Keep failures for 7 days
+});
+```
+
+### Bulk Add
+
+```typescript
+await emailQueue.addBulk([
+  { name: 'welcome', data: { userId: 'u_1' } },
+  { name: 'welcome', data: { userId: 'u_2' } },
+  { name: 'welcome', data: { userId: 'u_3' } },
+]);
+```
+
+### Repeatable Jobs (Cron / Interval)
+
+```typescript
+// Cron-based
+await emailQueue.add('daily-digest', {}, {
+  repeat: { pattern: '0 9 * * *' }, // Every day at 09:00 UTC
+});
+
+// Interval-based
+await emailQueue.add('health-check', {}, {
+  repeat: { every: 30_000 }, // Every 30 seconds
+});
+
+// Remove a repeatable job
+const repeatableJobs = await emailQueue.getRepeatableJobs();
+await emailQueue.removeRepeatableByKey(repeatableJobs[0].key);
+```
+
+## Worker — Processing Jobs
+
+```typescript
+import { Worker, Job } from 'bullmq';
+
+const worker = new Worker('email', async (job: Job) => {
+  // job.name — the job name ('welcome', 'invoice')
+  // job.data — the payload
+  // job.id   — unique job ID
+  // job.attemptsMade — current attempt number
+
+  await sendEmail(job.data.userId, job.data.template);
+
+  // Return value is stored as job result
+  return { sent: true, timestamp: Date.now() };
+}, {
+  connection,
+  concurrency: 5,            // Process up to 5 jobs in parallel
+  limiter: {
+    max: 100,                 // Max 100 jobs
+    duration: 60_000,         // Per 60 seconds (rate limiting)
+  },
+});
+```
+
+### Named Processors
+
+Handle different job types in one worker:
+
+```typescript
+const worker = new Worker('notifications', async (job: Job) => {
+  switch (job.name) {
+    case 'email':
+      return await handleEmail(job.data);
+    case 'sms':
+      return await handleSms(job.data);
+    case 'push':
+      return await handlePush(job.data);
+    default:
+      throw new Error(`Unknown job type: ${job.name}`);
+  }
+}, { connection });
+```
+
+### Progress Reporting
+
+```typescript
+const worker = new Worker('video-encode', async (job: Job) => {
+  for (let i = 0; i < 100; i++) {
+    await encodeChunk(i);
+    await job.updateProgress(i + 1);
+  }
+  return { encoded: true };
+}, { connection });
+```
+
+### Sandboxed Processors
+
+Run processor in a separate child process to isolate crashes:
+
+```typescript
+import { Worker } from 'bullmq';
+import path from 'node:path';
+
+const worker = new Worker(
+  'heavy-compute',
+  path.join(__dirname, 'processors/heavy-compute.js'),
+  { connection, useWorkerThreads: true } // Worker threads instead of child process
+);
+```
+
+```typescript
+// processors/heavy-compute.js
+import { SandboxedJob } from 'bullmq';
+
+export default async function (job: SandboxedJob) {
+  // Runs in isolation — a crash here does not kill the main process
+  return await computeResult(job.data);
+}
+```
+
+## Flow Producer — Parent-Child Job Graphs
+
+Create dependency trees where children must complete before the parent processes:
+
+```typescript
+import { FlowProducer } from 'bullmq';
+
+const flow = new FlowProducer({ connection });
+
+await flow.add({
+  name: 'generate-report',
+  queueName: 'reports',
+  data: { reportId: 'r_1' },
+  children: [
+    {
+      name: 'fetch-sales',
+      queueName: 'data-fetch',
+      data: { source: 'sales' },
+    },
+    {
+      name: 'fetch-inventory',
+      queueName: 'data-fetch',
+      data: { source: 'inventory' },
+    },
+  ],
+});
+// Parent job waits until both children complete, then processes with their results
+```
+
+Access child results from parent:
+
+```typescript
+const worker = new Worker('reports', async (job: Job) => {
+  const childValues = await job.getChildrenValues();
+  // childValues is Record<string, ReturnType> keyed by `queueName:jobId`
+  return buildReport(childValues);
+}, { connection });
+```
+
+## QueueEvents — Lifecycle Monitoring
+
+```typescript
+import { QueueEvents } from 'bullmq';
+
+const events = new QueueEvents('email', { connection });
+
+events.on('completed', ({ jobId, returnvalue }) => {
+  console.log(`Job ${jobId} completed with`, returnvalue);
+});
+
+events.on('failed', ({ jobId, failedReason }) => {
+  console.error(`Job ${jobId} failed:`, failedReason);
+});
+
+events.on('progress', ({ jobId, data: progress }) => {
+  console.log(`Job ${jobId} progress:`, progress);
+});
+
+events.on('stalled', ({ jobId }) => {
+  console.warn(`Job ${jobId} stalled — worker may have crashed`);
+});
+
+events.on('delayed', ({ jobId, delay }) => {
+  console.log(`Job ${jobId} delayed by ${delay}ms`);
+});
+```
+
+## Stalled Jobs
+
+A job is stalled when the worker processing it stops sending heartbeats (crash, memory pressure, blocked event loop). BullMQ automatically moves stalled jobs back to the waiting state.
+
+```typescript
+const worker = new Worker('critical', processor, {
+  connection,
+  stalledInterval: 30_000,   // Check for stalled jobs every 30s (default)
+  maxStalledCount: 1,        // Mark as failed after 1 stall (default: 1)
+  lockDuration: 30_000,      // How long a job lock lasts before considered stalled
+});
+```
+
+**Anti-pattern**: Setting `lockDuration` too low for long-running jobs. If your job takes 5 minutes, set `lockDuration` to at least 600_000 (10 minutes) to avoid false stalls.
+
+## Dashboard — Bull Board
+
+```bash
+pnpm add @bull-board/api @bull-board/express
+```
+
+```typescript
+import { createBullBoard } from '@bull-board/api';
+import { BullMQAdapter } from '@bull-board/api/bullMQAdapter';
+import { ExpressAdapter } from '@bull-board/express';
+import express from 'express';
+
+const serverAdapter = new ExpressAdapter();
+serverAdapter.setBasePath('/admin/queues');
+
+createBullBoard({
+  queues: [
+    new BullMQAdapter(emailQueue),
+    new BullMQAdapter(videoQueue),
+  ],
+  serverAdapter,
+});
+
+const app = express();
+app.use('/admin/queues', serverAdapter.getRouter());
+app.listen(3000);
+```
+
+## Testing
+
+```typescript
+import { Queue, Worker, Job } from 'bullmq';
+import { describe, it, expect, afterAll } from 'vitest';
+
+describe('email queue', () => {
+  const connection = { host: '127.0.0.1', port: 6379 };
+  const queue = new Queue('test-email', { connection });
+
+  afterAll(async () => {
+    await queue.obliterate({ force: true }); // Clean up test queue
+    await queue.close();
+  });
+
+  it('processes welcome email', async () => {
+    const result = await new Promise<string>(async (resolve) => {
+      const worker = new Worker('test-email', async (job: Job) => {
+        return `sent-to-${job.data.userId}`;
+      }, { connection });
+
+      worker.on('completed', async (job: Job) => {
+        resolve(await job.returnvalue);
+        await worker.close();
+      });
+
+      await queue.add('welcome', { userId: 'u_test' });
+    });
+
+    expect(result).toBe('sent-to-u_test');
+  });
+});
+```
+
+**Testing anti-pattern**: Do not mock BullMQ internals. Use a real Redis instance (Docker or CI service) for integration tests. Mock only external services your processor calls.
+
+## Graceful Shutdown
+
+```typescript
+import { Worker, Queue } from 'bullmq';
+
+const workers: Worker[] = [];
+
+function createWorker(name: string, processor: (job: Job) => Promise<unknown>) {
+  const worker = new Worker(name, processor, { connection });
+  workers.push(worker);
+  return worker;
+}
+
+async function shutdown() {
+  console.log('Shutting down gracefully...');
+
+  // Close workers first — finishes current jobs, stops taking new ones
+  await Promise.all(workers.map((w) => w.close()));
+
+  // Then close queues
+  await emailQueue.close();
+
+  // Finally close Redis connection
+  await connection.quit();
+
+  process.exit(0);
+}
+
+process.on('SIGTERM', shutdown);
+process.on('SIGINT', shutdown);
+```
+
+## Anti-Patterns
+
+| Anti-Pattern | Why It Fails | Correct Approach |
+|-------------|-------------|-----------------|
+| Storing large payloads in job data | Redis memory pressure, slow serialization | Store data externally, pass IDs in job data |
+| Not setting `removeOnComplete` | Redis fills up with completed job records | Use `{ count: 1000 }` or `{ age: 86400 }` |
+| Using `maxRetriesPerRequest` on worker connections | BullMQ blocking commands fail | Set `maxRetriesPerRequest: null` |
+| Processing side effects without idempotency | Retries cause duplicate actions | Design processors to be idempotent |
+| Not closing workers on shutdown | Jobs get stalled, then retried | Call `worker.close()` on SIGTERM/SIGINT |
+| Short `lockDuration` with long jobs | Jobs falsely stall mid-processing | Set `lockDuration` > max expected job duration |
+
+## Decision Guide
+
+| Need | BullMQ Feature |
+|------|---------------|
+| Retry failed jobs | `attempts` + `backoff` options |
+| Schedule future work | `delay` option |
+| Recurring tasks | `repeat` with cron pattern or interval |
+| Rate limit processing | `limiter` on Worker |
+| Job dependencies | FlowProducer parent-child trees |
+| Crash isolation | Sandboxed processors |
+| Monitor job status | QueueEvents or Bull Board dashboard |
+| Priority processing | `priority` option (lower = higher priority) |