---

name: job-queue description: >- Build job queues and background worker systems using BullMQ, Celery, or Sidekiq. Use when you need to offload slow tasks from request handlers — email sending, PDF generation, image processing, data exports, or any work that takes more than a few hundred milliseconds. Covers job priorities, concurrency control, scheduled jobs, progress tracking, and graceful shutdown. Trigger words: background job, worker, queue, async task, BullMQ, Celery, cron job, scheduled task, job retry. license: Apache-2.0 compatibility: "Redis 7+ for BullMQ/Sidekiq. RabbitMQ for Celery. Node.js 18+ or Python 3.10+." metadata: author: terminal-skills version: "1.0.0" category: development tags: ["job-queue", "background-workers", "async", "BullMQ"]

Job Queue

Overview

This skill helps you build production-grade background job processing systems. It covers queue architecture, worker concurrency, job priorities, retry strategies, scheduled/recurring jobs, progress reporting, and graceful shutdown. The patterns work across BullMQ (Node.js), Celery (Python), and Sidekiq (Ruby).

Instructions

1. Set up the queue and define job types

Create typed job definitions and a queue instance:

// src/jobs/types.ts
export interface JobMap {
  "email:send": { to: string; template: string; data: Record<string, string> };
  "pdf:generate": { reportId: string; format: "a4" | "letter" };
  "export:csv": { userId: string; query: string; columns: string[] };
  "image:resize": { sourceUrl: string; widths: number[] };
}

// src/jobs/queues.ts
import { Queue } from "bullmq";
import { JobMap } from "./types";

const connection = { host: "localhost", port: 6379 };

export const emailQueue = new Queue<JobMap["email:send"]>("email", { connection });
export const pdfQueue = new Queue<JobMap["pdf:generate"]>("pdf", { connection });
export const exportQueue = new Queue<JobMap["export:csv"]>("export", { connection });
export const imageQueue = new Queue<JobMap["image:resize"]>("image", { connection });

2. Implement workers with concurrency control

// src/workers/email-worker.ts
import { Worker, Job } from "bullmq";
import { JobMap } from "../jobs/types";

const emailWorker = new Worker<JobMap["email:send"]>(
  "email",
  async (job: Job) => {
    const { to, template, data } = job.data;
    await job.updateProgress(10);
    const html = await renderTemplate(template, data);
    await job.updateProgress(50);
    await sendEmail(to, html);
    await job.updateProgress(100);
    return { sentAt: new Date().toISOString() };
  },
  {
    connection: { host: "localhost", port: 6379 },
    concurrency: 10,        // Process 10 emails in parallel
    limiter: { max: 100, duration: 60000 }, // Rate limit: 100/minute
  }
);

emailWorker.on("completed", (job) => {
  console.log(`Email sent: job ${job.id} → ${job.data.to}`);
});

emailWorker.on("failed", (job, err) => {
  console.error(`Email failed: job ${job?.id} — ${err.message}`);
});

3. Add job scheduling and priorities

// Delayed job — send welcome email 30 minutes after signup
await emailQueue.add("email:send", {
  to: "newuser@example.com",
  template: "welcome",
  data: { name: "Alex" },
}, { delay: 30 * 60 * 1000 });

// Priority jobs — password resets jump the queue
await emailQueue.add("email:send", {
  to: "user@example.com",
  template: "password-reset",
  data: { resetLink: "https://app.example.com/reset/abc123" },
}, { priority: 1 }); // Lower number = higher priority

// Recurring job — daily digest at 8:00 AM UTC
await emailQueue.add("email:send", {
  to: "digest",
  template: "daily-digest",
  data: {},
}, {
  repeat: { pattern: "0 8 * * *" },
  jobId: "daily-digest", // Prevent duplicates
});

4. Implement graceful shutdown

// src/workers/shutdown.ts
const workers = [emailWorker, pdfWorker, exportWorker, imageWorker];

async function gracefulShutdown(signal: string): Promise<void> {
  console.log(`Received ${signal}. Closing workers gracefully...`);
  await Promise.all(workers.map((w) => w.close()));
  console.log("All workers closed. Exiting.");
  process.exit(0);
}

process.on("SIGTERM", () => gracefulShutdown("SIGTERM"));
process.on("SIGINT", () => gracefulShutdown("SIGINT"));

Examples

Example 1: PDF report generation queue

Prompt: "Build a background job system for generating PDF reports. Users request a report, get a job ID back immediately, and can poll for progress. Reports take 10-30 seconds to generate."

Agent output:

• Creates src/jobs/pdf-queue.ts with typed job definitions
• Creates src/workers/pdf-worker.ts with progress updates at each stage (query data → format → render → upload)
• Creates src/routes/reports.ts with POST /reports (enqueue, return job ID) and GET /reports/:jobId/status (return progress percentage and download URL when complete)
• Adds retry logic: 3 attempts with 10-second backoff

Example 2: Image processing pipeline

Prompt: "I need to process uploaded images: resize to 3 widths (200, 800, 1600px), convert to WebP, and upload to cloud storage. Handle up to 500 images per hour."

Agent output:

• Creates src/workers/image-worker.ts with sharp-based resize and conversion pipeline
• Sets concurrency to 4 (CPU-bound work, matches core count)
• Adds per-image progress tracking (useful for batch uploads)
• Creates src/jobs/image-pipeline.ts with a flow: resize → convert → upload as chained jobs

Guidelines

• Keep jobs serializable — job data must survive JSON round-trips. Pass IDs and URLs, not buffers or streams.
• Set appropriate concurrency — CPU-bound work (image processing): match core count. I/O-bound (email, API calls): 10-50 concurrent.
• Always implement graceful shutdown — SIGTERM should let running jobs finish before the process exits.
• Use job IDs for idempotency — set a deterministic jobId to prevent the same job from being enqueued twice.
• Monitor queue depth — a growing queue means workers can't keep up. Alert when backlog exceeds 5 minutes of processing time.
• Separate queues by workload type — don't let a slow PDF generation block fast email sends.
• Store results externally — BullMQ job results are cleaned up by default. Persist important results in your database.