How do you run robust background jobs (Celery/ARQ)?

Question

Accepted Answer

For **robust, durable background jobs** (beyond FastAPI's lightweight `BackgroundTasks`), you use a real task queue: **Celery** (the established standard) or **ARQ** (a modern async-native queue). These run heavy, retryable, scheduled work in separate worker processes, backed by a broker like Redis.

## Why not just BackgroundTasks?

```text
BackgroundTasks runs in the web process → NO persistence (lost on crash), NO retries,
NO scheduling, and heavy work ties up the worker. Fine only for quick fire-and-forget.
For durable/critical/heavy/scheduled jobs → use a real task queue (Celery / ARQ).
```

## Celery — define and queue tasks

```python
# tasks.py
from celery import Celery
celery_app = Celery("tasks", broker="redis://localhost:6379")

@celery_app.task(bind=True, max_retries=3)
def process_video(self, video_id: int):
    try:
        do_heavy_processing(video_id)        # runs in a WORKER, not the web process
    except Exception as exc:
        raise self.retry(exc=exc, countdown=60)   # automatic retry on failure
```

```python
# in a FastAPI endpoint — QUEUE the task, return immediately
@app.post("/videos/{id}/process")
def process(id: int):
    process_video.delay(id)                  # enqueue → returns at once (doesn't block)
    return {"status": "processing started"}
# run workers separately: celery -A tasks worker
```

The endpoint enqueues the job and responds instantly; a separate **worker** process executes it with persistence and retries.

## ARQ — async-native (fits FastAPI's async model)

```python
# ARQ is built on asyncio — a natural fit for async FastAPI codebases
async def process_video(ctx, video_id: int):
    await do_async_processing(video_id)

# enqueue from an endpoint
await redis.enqueue_job("process_video", video_id)
```

ARQ is a lighter, async-first alternative — appealing when your codebase is already async.

## What a real task queue provides

```text
✓ Persistence — jobs survive crashes/restarts (stored in the broker)
✓ Retries with backoff — resilience against transient failures
✓ Scheduling — periodic/cron-like jobs (Celery Beat)
✓ Distribution & scaling — workers scale independently of the web servers
✓ Concurrency & rate limiting, task routing/priorities, result tracking
```

## Why it matters

Running robust background jobs is essential for building production-grade FastAPI applications that handle heavy or deferred work reliably — and understanding when and how to use a real task queue is important senior-level knowledge.

The core judgment is recognizing the **limits of `BackgroundTasks`** (no persistence, retries, or scheduling — fine only for quick, non-critical work) and reaching for **Celery or ARQ** when work must be durable, retryable, scheduled, heavy, or distributed (processing files/videos, sending bulk emails, data syncing, generating reports).

These queues offer critical production capabilities: **persistence** (jobs survive crashes via the broker, so work isn't lost), **automatic retries** (resilience against transient failures), **scheduling** (periodic jobs), and **independent scaling** of workers from web servers.

Knowing how to define and enqueue tasks (Celery's `.delay()`, ARQ's async-native API), the architecture (broker + workers), and especially the **decision of when to use a real queue versus BackgroundTasks**, is valuable for architecting reliable applications that offload heavy work without losing it — a common, important requirement that distinguishes robust production systems from fragile ones.