python - celery : launch chord callback after its associated body-6ren

python - celery : launch chord callback after its associated body

转载作者：太空宇宙更新时间：2023-11-04 02:38:53

当我启动包含一组任务和一个回调的 chord() 列表时，只有在完成所有任务组后才会调用回调，即使是不在当前和弦。

这里是更好解释的代码:

import time

from celery import Celery, group, chord

app = Celery('tasks')
app.config_from_object('celeryconfig')

@app.task(name='SHORT_TASK')
def short_task(t):
    time.sleep(t)
    return t

@app.task(name='FINISH_GROUP')
def finish_group(res, nb):
    print("Pipe #{} finished".format(nb))
    return True

@app.task
def main(total):
    tasks = []
    for nb in range(1, total+1):
        short_tasks = [short_task.si(i) for i in [0.5, 0.75, 1]]

        chord(
            group(short_tasks),
            finish_group.s(nb)
        ).apply_async()

例如，我用 5 个项目启动它:

In [5]: main.delay(5)
Out[5]: <AsyncResult: db1f97f0-ff7a-4651-b2f9-11e27a001480>

结果:

[2017-11-06 13:50:38,374: INFO/MainProcess] Received task: tasks.main[6da738b5-4eae-4de4-9ac5-1dc67d210f1d]  
[2017-11-06 13:50:38,409: INFO/MainProcess] Received task: SHORT_TASK[9581f9e0-1128-4b87-ae6b-16f238b2337e]  
[2017-11-06 13:50:38,411: INFO/MainProcess] Received task: SHORT_TASK[579dc498-3770-4385-a25a-06173fbe639c]  
[2017-11-06 13:50:38,412: INFO/MainProcess] Received task: SHORT_TASK[bfafb943-46d8-42e3-941f-b48a9c8e0186]  
[2017-11-06 13:50:38,414: INFO/MainProcess] Received task: SHORT_TASK[a1208f06-250f-48ac-b3df-45c4525fe8eb]  
[2017-11-06 13:50:38,416: INFO/MainProcess] Received task: SHORT_TASK[86ee7408-9d61-4909-bce8-c42cf691e9c2]  
[2017-11-06 13:50:38,416: INFO/MainProcess] Received task: SHORT_TASK[e2bb22c0-1d20-4da7-91d9-45b7ed8bfc6f]  
[2017-11-06 13:50:38,419: INFO/MainProcess] Received task: SHORT_TASK[7551199b-4690-45dd-a434-3911861f0093]  
[2017-11-06 13:50:38,420: INFO/MainProcess] Received task: SHORT_TASK[362d18f4-2252-4a31-ad21-4a2d192fd22e]  
[2017-11-06 13:50:38,421: INFO/MainProcess] Received task: SHORT_TASK[7561c33b-7020-4feb-b054-3919e4ae31c2]  
[2017-11-06 13:50:38,423: INFO/MainProcess] Received task: SHORT_TASK[3ac997f5-6d0f-43b6-ab15-a6827a26665f]  
[2017-11-06 13:50:38,423: INFO/MainProcess] Received task: SHORT_TASK[8b2ebb3a-293c-4bb8-88a3-5242750a082e]  
[2017-11-06 13:50:38,423: INFO/ForkPoolWorker-3] Task tasks.main[6da738b5-4eae-4de4-9ac5-1dc67d210f1d] succeeded in 0.048569423001026735s: None
[2017-11-06 13:50:38,424: INFO/MainProcess] Received task: SHORT_TASK[efd68688-ec4a-418d-83b9-a55fd6cc1541]  
[2017-11-06 13:50:38,427: INFO/MainProcess] Received task: SHORT_TASK[4e82540e-f935-4288-828f-c6f66f84139a]  
[2017-11-06 13:50:38,427: INFO/MainProcess] Received task: SHORT_TASK[a94e4ec4-adcb-4a0f-b184-b36650105ed5]  
[2017-11-06 13:50:38,427: INFO/MainProcess] Received task: SHORT_TASK[0d4b5e24-7aaa-4eb8-8e54-70e769dfdb39]  
[2017-11-06 13:50:38,918: INFO/ForkPoolWorker-2] Task SHORT_TASK[9581f9e0-1128-4b87-ae6b-16f238b2337e] succeeded in 0.5051485379808582s: 0.5
[2017-11-06 13:50:38,926: INFO/ForkPoolWorker-3] Task SHORT_TASK[a1208f06-250f-48ac-b3df-45c4525fe8eb] succeeded in 0.5012409449846018s: 0.5
[2017-11-06 13:50:39,165: INFO/ForkPoolWorker-1] Task SHORT_TASK[579dc498-3770-4385-a25a-06173fbe639c] succeeded in 0.7524393269850407s: 0.75
[2017-11-06 13:50:39,445: INFO/ForkPoolWorker-4] Task SHORT_TASK[bfafb943-46d8-42e3-941f-b48a9c8e0186] succeeded in 1.031865488999756s: 1
[2017-11-06 13:50:39,448: INFO/MainProcess] Received task: FINISH_GROUP[4506631f-f9cc-4e9e-a9e7-9a59c8f7c998]  
[2017-11-06 13:50:39,668: INFO/ForkPoolWorker-1] Task SHORT_TASK[7551199b-4690-45dd-a434-3911861f0093] succeeded in 0.501304400007939s: 0.5
[2017-11-06 13:50:39,672: INFO/ForkPoolWorker-2] Task SHORT_TASK[86ee7408-9d61-4909-bce8-c42cf691e9c2] succeeded in 0.7513346789928619s: 0.75
[2017-11-06 13:50:39,932: INFO/ForkPoolWorker-3] Task SHORT_TASK[e2bb22c0-1d20-4da7-91d9-45b7ed8bfc6f] succeeded in 1.0058077470166609s: 1
[2017-11-06 13:50:39,936: INFO/MainProcess] Received task: FINISH_GROUP[f143d581-799b-45ff-9e11-edc0bb88006a]  
[2017-11-06 13:50:40,175: INFO/ForkPoolWorker-2] Task SHORT_TASK[3ac997f5-6d0f-43b6-ab15-a6827a26665f] succeeded in 0.502920284983702s: 0.5
[2017-11-06 13:50:40,198: INFO/ForkPoolWorker-4] Task SHORT_TASK[362d18f4-2252-4a31-ad21-4a2d192fd22e] succeeded in 0.752579735009931s: 0.75
[2017-11-06 13:50:40,685: INFO/ForkPoolWorker-3] Task SHORT_TASK[8b2ebb3a-293c-4bb8-88a3-5242750a082e] succeeded in 0.7518302960088477s: 0.75
[2017-11-06 13:50:40,701: INFO/ForkPoolWorker-4] Task SHORT_TASK[4e82540e-f935-4288-828f-c6f66f84139a] succeeded in 0.5013290829956532s: 0.5
[2017-11-06 13:50:40,715: INFO/ForkPoolWorker-1] Task SHORT_TASK[7561c33b-7020-4feb-b054-3919e4ae31c2] succeeded in 1.0464465210097842s: 1
[2017-11-06 13:50:40,715: WARNING/ForkPoolWorker-1] Pipe #1 finished
[2017-11-06 13:50:40,716: INFO/ForkPoolWorker-1] Task FINISH_GROUP[4506631f-f9cc-4e9e-a9e7-9a59c8f7c998] succeeded in 0.000513697013957426s: True
[2017-11-06 13:50:40,716: WARNING/ForkPoolWorker-1] Pipe #2 finished
[2017-11-06 13:50:40,717: INFO/ForkPoolWorker-1] Task FINISH_GROUP[f143d581-799b-45ff-9e11-edc0bb88006a] succeeded in 0.0003622350050136447s: True
[2017-11-06 13:50:40,718: INFO/MainProcess] Received task: FINISH_GROUP[fc9be8c2-99f7-46b2-a810-47023e0a072a]  
[2017-11-06 13:50:40,718: WARNING/ForkPoolWorker-1] Pipe #3 finished
[2017-11-06 13:50:40,718: INFO/ForkPoolWorker-1] Task FINISH_GROUP[fc9be8c2-99f7-46b2-a810-47023e0a072a] succeeded in 0.00038264598697423935s: True
[2017-11-06 13:50:41,215: INFO/ForkPoolWorker-2] Task SHORT_TASK[efd68688-ec4a-418d-83b9-a55fd6cc1541] succeeded in 1.0379863310081419s: 1
[2017-11-06 13:50:41,219: INFO/MainProcess] Received task: FINISH_GROUP[6a4dc66e-2232-4bad-9d85-9fbc63b8b847]  
[2017-11-06 13:50:41,221: WARNING/ForkPoolWorker-2] Pipe #4 finished
[2017-11-06 13:50:41,222: INFO/ForkPoolWorker-2] Task FINISH_GROUP[6a4dc66e-2232-4bad-9d85-9fbc63b8b847] succeeded in 0.0018843600118998438s: True
[2017-11-06 13:50:41,440: INFO/ForkPoolWorker-3] Task SHORT_TASK[a94e4ec4-adcb-4a0f-b184-b36650105ed5] succeeded in 0.7531412789830938s: 0.75
[2017-11-06 13:50:41,708: INFO/ForkPoolWorker-4] Task SHORT_TASK[0d4b5e24-7aaa-4eb8-8e54-70e769dfdb39] succeeded in 1.005872479028767s: 1
[2017-11-06 13:50:41,711: INFO/MainProcess] Received task: FINISH_GROUP[388ee1c3-b80c-41af-bbfd-29b968e90aff]  
[2017-11-06 13:50:41,712: WARNING/ForkPoolWorker-3] Pipe #5 finished
[2017-11-06 13:50:41,712: INFO/ForkPoolWorker-3] Task FINISH_GROUP[388ee1c3-b80c-41af-bbfd-29b968e90aff] succeeded in 0.0005500270053744316s: True

我启动了一个并发度为 4(prefork)的 Celery。

我们可以看到一开始收到了 15 个 SHORT_TASK，然后 worker 执行它，然后才调用 FINISH_GROUP 任务。

是否可以在他们相关的 SHORT_TASK 完成后启动 FINISH_GROUP 任务，而不是等待所有其他不相关的 SHORT_TASK？

也许我的 Canvas 不正确，或者是错误的 Celery 配置，我不知道。

感谢您的帮助!

最佳答案

你的测试有偏差，因为你只使用了一个 worker，time.sleep() 会阻塞那个 worker。这意味着即使并发为 4，它也不再处理任务。

Is it possible to launch the FINISH_GROUP task just after their associated SHORT_TASK have been finished, and not wait all the others non-related SHORT_TASK please ?

目前您没有等待其他short_task 完成，它们都安排在同一时间执行。由于您正在使用 sleep ，因此一旦所有 short_task 各自的和弦结束，finish_group 就会被调用。

您当前的执行方式如下:

| chord 1      | chord 2      | chord 3      |
|--------------|--------------|--------------|
| short_task 1 |              |              |      |
|              | short_task 1 |              |      |
|              |              | short_task 1 |      |
| short_task 2 |              |              |      |
|              | short_task 2 |              |      |
|              |              | short_task 2 |      |
| short_task 3 |              |              |      v
|              | short_task 3 |              | execution order
|              |              | short_task 3 |
| finish_group |              |              |
|              | finish_group |              |
|              |              | finish_group |

如果删除 sleep ，添加更多工作人员，或使用 gevent。它应该看起来像这样:

| chord 1          | chord 2          | chord 3          |
|------------------|------------------|------------- ----|
| short_task 1,2,3 | short_task 1,2,3 | short_task 1,2,3 |
| finish_group     | finish_group     | finish_group     |

而且您应该会看到在同一行上的任务将以不同的顺序出现在日志中(具体取决于哪个工作人员先处理)。但是 finish_group 仍然是最后一个。

请注意，使用 chord 时，不需要group您的任务

chord(
    short_tasks,
    finish_group.s(nb)
)

相同的代码，但使用了 gevent:

import gevent
from celery import Celery, group, chord, chain

app = Celery('tasks', broker='redis://localhost/4', backend='redis://localhost/5')


@app.task()
def short_task(nb, i):
    print('TEST: start short_task({}, {})'.format(nb, i))
    gevent.sleep(1)
    print('TEST: end   short_task({}, {})'.format(nb, i))
    return i


@app.task(name='FINISH_GROUP')
def finish_group(results, nb):
    print('TEST: finish_group({}) -> {}'.format(nb, results))


@app.task
def main(total):
    for nb in range(1, total+1):
        short_tasks = [short_task.si(nb, i) for i in range(3)]

        chord(short_tasks, finish_group.s(nb)).apply_async()

启动:

$ celery worker -A celery_test --loglevel=debug --concurrency=20 -P gevent 2>&1 | grep TEST

由于并行执行，输出将是困惑的。

[2017-11-06 16:40:08,085] TEST: start short_task(1, 0)
[2017-11-06 16:40:08,088] TEST: start short_task(1, 1)
[2017-11-06 16:40:08,091] TEST: start short_task(1, 2)
[2017-11-06 16:40:08,092] TEST: start short_task(2, 0)
[2017-11-06 16:40:08,094] TEST: start short_task(2, 1)
[2017-11-06 16:40:08,096] TEST: start short_task(2, 2)
[2017-11-06 16:40:08,100] TEST: start short_task(3, 0)
[2017-11-06 16:40:08,101] TEST: start short_task(3, 1)
[2017-11-06 16:40:08,103] TEST: start short_task(3, 2)
# ^ all short_task have been started at the same time

[2017-11-06 16:40:09,085] TEST: end   short_task(1, 0)
[2017-11-06 16:40:09,089] TEST: end   short_task(1, 1)
[2017-11-06 16:40:09,093] TEST: end   short_task(1, 2)
[2017-11-06 16:40:09,106] TEST: end   short_task(2, 0)
[2017-11-06 16:40:09,106] TEST: end   short_task(2, 1)
[2017-11-06 16:40:09,107] TEST: end   short_task(2, 2)
[2017-11-06 16:40:09,107] TEST: end   short_task(3, 0)
[2017-11-06 16:40:09,108] TEST: end   short_task(3, 1)
[2017-11-06 16:40:09,108] TEST: end   short_task(3, 2)
# ^ total execution is only 1 second since 9 greenlet have slept together

[2017-11-06 16:40:09,115] TEST: finish_group(1) -> [0, 1, 2]
[2017-11-06 16:40:09,126] TEST: finish_group(2) -> [2, 1, 0]
[2017-11-06 16:40:09,128] TEST: finish_group(3) -> [0, 1, 2]
# ^ order of results are mixed depending of which greenlet finished first

关于python - celery : launch chord callback after its associated body，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/47137592/

文章推荐： Python/Beautiful soup find_all() 找不到全部

文章推荐： C 读取文件和打印缓冲区

文章推荐： c - (C) 在多线程客户端和服务器中使用互斥量

文章推荐： python - Flask-security: "current_user"是从哪里来的？

java - RxJava : what is difference between callbacks in doOnError ('callback' ) and subscribe(*, 'callback')
在我的上一个项目中，我使用了 rxJava，我意识到 observable.doOnError('onErrorCallback').subscribe(action) 和 observable.su
c++ - 为什么 (*callback)() 有效，而 *callback() 或 *callback 在 C++ 中不起作用
我是一名 C++ 初学者，我认为要真正学习指针和引用，我应该尝试创建一个回调函数，这是我在 JavaScript 中认为理所当然的事情。但是，对于我的一生，我不知道为什么这些括号在 (*callba
callback - 达特朗 : How to implemet a callback method
我在库中有一个类，它具有在事件发生时执行的“onMessage”方法。 OnMessage 在执行时需要调用属于主应用程序中的类的“回调”方法。我假设这将通过构造函数完成，但我不知道它是如何实现的。
jquery - $.Callbacks().disable() 与 $.Callbacks().lock()
两者的 jQuery 文档基本上说明了相同的事情，所以我想知道两者之间是否有任何重大差异(如果有的话)。谢谢! 最佳答案这方面的文档实际上非常糟糕，所以这是我在 studying the sourc
javascript - callback && callback() 在 javascript 中是什么意思
这个问题在这里已经有了答案: Using &&'s short-circuiting as an if statement? (6 个答案) Omitting the second expressi
callback - 戈朗 : evaluate variable in callback declaration
我正在尝试在 golang 中定义一个回调: package main func main() { x, y := "old x ", "old y" callback
callback - 谷歌图表 API : adding arguments to existing callback
我有一个页面，其中包含从 Google 电子表格生成的许多图表。典型代码如下所示: var url = "http://my.googlespreadsheet.com/tq?argumentshe
callback - 订阅已弃用 : Use an observer instead of an error callback
当我运行 linter 时，它显示: subscribe is deprecated: Use an observer instead of an error callback 代码来自 this a
c# - Callback 与 new AsyncCallback(Callback) 有什么不同？
对于异步套接字 // accept ... listener.BeginAccept( new AsyncCallback(AcceptCallback), listener); // listene
javascript - 根据 callback(true)/callback(false) 执行操作
我希望能够根据在前面的函数中调用的是 callback(true) 还是 callback(false) 在回调函数中执行一些逻辑。示例: foo.doFunction = function (pa
javascript - jQuery:以下代码块中比较的含义是什么 - callback && function() {callback.call();}
从 jQuery.scrollTo.js 库中看到这个 block (在 v1.4 的第 184 行)。 function animate( callback ){ $elem.animate
android - "callback(value)"和 "callback.invoke(value)"有什么区别？
我正在尝试在我的应用中使用一些回调，它与 "callback(value)" 和 "callback.invoke(value)" 一起工作正确调用回调。我想知道“回调(值)”是否只是一个缩短版本，
python - 访问 tf.keras.callbacks.Callback 中已弃用的属性 "validation_data"
我决定从 keras 切换到 tf.keras(建议使用 here)。因此我安装了 tf.__version__=2.0.0和 tf.keras.__version__=2.2.4-tf .在我的旧版
javascript - nodejs return callback() 和 just callback() 有什么区别
我认为这实际上可能会回答我关于 Stack Overflow 的另一个问题如果我能确认这一点。返回回调和只调用回调有什么区别？我看到代码执行其中之一/或/两者，并试图思考为什么以及何时执行哪个。
callback - 如何 : Idiomatic Rust for callbacks with gtk (rust-gnome)
我目前正在学习 Rust 并希望用它来开发 GUI基于 GTK+ 的应用程序。我的问题与注册回调有关在这些回调中响应 GTK 事件/信号和变异状态。我有一个有效但不优雅的解决方案，所以我想问一下是否有
javascript - React_Redux : Pass parameter in callback during continually callback function
我在回调函数中传递参数时遇到问题。我使用 redux-form，当我更改 SkinList 中的选择时，它会触发 onChange 回调 - activeSkinChange 方法在activeSk
javascript - Node : Dealing with Multiple Promise callbacks (Callback hell)
我有 8 个相互依赖的回调。我的想法是要有一个更具可读性的过程，但我不明白如何处理这个问题。我的回调 hell 的一个例子是: return new Promise(function (resolv
How do I use the result of a callback and pass that result to the next callback?(如何使用回调的结果并将该结果传递给下一个回调？)
因此，我的函数接受一个值和任意数量的回调作为参数(我应该使用扩散操作符吗？)该函数应该返回通过所有给定回调传递该值的最终结果。。我返回的“CB2(Res1)”不是一个函数。如何将第一个回调的结果传递给
callback - Vert.x 是 "based on callbacks"(而不是 future )是什么意思？
在谈到 future 和回调时，documentation说是 The Vert.x core APIs are based on callbacks to notify of asynchronou
Facebook 连接演示 - 是 "Callback URL"== "Connect Callback URL"吗？
我开始觉得自己很蠢。我正在关注 Facebook-Connect 演示“The Run Around”。当我导航到 http://www.[mysite].com/testing/register_

太空宇宙

个人简介

我是一名优秀的程序员,十分优秀！

作者热门文章

滴滴打车优惠券免费领取

全站热门文章

首页

博学

6Ren·AI

商城

python - celery : launch chord callback after its associated body