narq 0.2.0

A distributed task queue built with asyncio and redis, with built-in web interface

Introduction

narq is a distributed task queue with asyncio and redis, which is built upon ReArq (itself a rewrite of arq)

Motivations

This project is an independent fork of ReArq because it is fundamentally different in its goals. Narq is intended as a simple to reason about production-grade task queue.

Features

• AsyncIO support, easy integration with FastAPI.
• Delayed tasks, cron tasks and async task support.
• Full-featured built-in web interface.
• Built-in distributed task lock to ensure a given task is ran one at a time.
• Other powerful features to be discovered.

Web Interface

Requirements

• Redis >= 5.0

Quick Start

Task Definition

# main.py
from narq import Narq

narq = Narq(db_url='mysql://root:123456@127.0.0.1:3306/narq')


@narq.on_shutdown
async def on_shutdown():
    # you can do some clean up work here like close db and so on...
    print("shutdown")


@narq.on_startup
async def on_startup():
    # you can do some initialization work here
    print("startup")


@narq.task(queue="q1")
async def add(self, a, b):
    return a + b


@narq.task(cron="*/5 * * * * * *")  # run task per 5 seconds
async def timer(self):
    return "timer"

Run narq worker

> narq main:narq worker -q q1 -q q2 # consume tasks from q1 and q2 as the same time

2021-03-29 09:54:50.464 | INFO     | narq.worker:_main:95 - Started worker successfully on queue: narq:queue:default
2021-03-29 09:54:50.465 | INFO     | narq.worker:_main:96 - Registered tasks: add, sleep, timer_add
2021-03-29 09:54:50.465 | INFO     | narq.worker:log_redis_info:86 - redis_version=6.2.1 mem_usage=1.43M clients_connected=5 db_keys=6

Run narq timer

If you have timing task or delay task, you should run another command also:

> narq main:narq timer

2021-03-29 09:54:43.878 | INFO     | narq.worker:_main:275 - Start timer successfully
2021-03-29 09:54:43.887 | INFO     | narq.worker:_main:277 - Registered timer tasks: timer_add
2021-03-29 09:54:43.894 | INFO     | narq.worker:log_redis_info:86 - redis_version=6.2.1 mem_usage=1.25M clients_connected=2 db_keys=6

Also, you can run timer with worker together by narq main:narq worker -t.

Integration with FastAPI

from fastapi import FastAPI

app = FastAPI()


@app.on_event("shutdown")
async def shutdown() -> None:
    await narq.close()


# then run task in view
@app.get("/test")
async def test():
    job = await add.delay(args=(1, 2))
    # or
    job = await add.delay(kwargs={"a": 1, "b": 2})
    # or
    job = await add.delay(1, 2)
    # or
    job = await add.delay(a=1, b=2)
    result = await job.result(timeout=5)  # wait result for 5 seconds
    print(result.result)
    return result

Start web interface

> narq main:narq server
Usage: narq server [OPTIONS]

  Start rest api server.

Options:
  --host TEXT         Listen host.  [default: 0.0.0.0]
  -p, --port INTEGER  Listen port.  [default: 8000]
  -h, --help          Show this message and exit..

After starting the server, check https://127.0.0.1:8000/docs to see all endpoints and https://127.0.0.1:8000 to use the web interface.

Other options will be passed into uvicorn directly, such as --root-path etc.

narq main:narq server --host 0.0.0.0 --root-path /narq

Mount as FastAPI sub app

If you have an existing FastAPI service, to simplify your deployment you might want to mount the narq server as a FastAPI sub app.

from fastapi import FastAPI

from examples.tasks import narq
from narq.server.app import app as narq

app = FastAPI()

app.mount("/narq", narq_app)
narq_app.set_narq(narq)

Start worker inside app

You can also start worker inside your app.

@app.on_event("startup")
async def startup():
    await narq.init()
    await narq.start_worker(with_timer=True, block=False)

ThanksTo

• arq, Fast job queuing and RPC in python with asyncio and redis.
• ReArq, Improved arq rewrite with an API + web interface

License

This project is licensed under the Apache-2.0 License.