Flower broker API is not working.

[rive-n]

Describe the bug I have infrastructure like: Celery worker + RabbitMQ as a broker + Redis as backend + Flower. So the problem is in that task are creating like once in 100 times i guess. And i can't use /broker in flower. My dockerfiles and docker-compose is below with trace: [E 211208 14:33:30 broker:79] RabbitMQ management API call failed: Timeout while connecting. By the way, in logs there is NOTHING about problems on flower or Rabbit side.

Dockerfile-flower:

FROM python:3.8-alpine

ENV PYTHONBUFFERED=0

# need to copy flower requirements #
# flower uses celery v5 #
RUN mkdir /flowerapp
WORKDIR /flowerapp
COPY requirements/flower-requirements.txt .
COPY entrypoints/flower-entrypoint.sh .
RUN pip install -r flower-requirements.txt

# creating user, no need to run from root! #
RUN addgroup -S flowerg && adduser -S flower -G flowerg
# USER flower

ENTRYPOINT /flowerapp/flower-entrypoint.sh

Flower-entrypoint:

#!/bin/sh

celery --broker amqp://$RABBITMQ_DEFAULT_USER:$RABBITMQ_DEFAULT_PASS@rabbit:5672// flower

With this configuration there are 2 types of errors. First - upper. And second is about {bad_header,<<"GET /api">>} - ref on https://github.com/mher/flower/issues/1036. So if I am able to fix bad_header, i can't fix Timeout.

So my final docker-compose with no Dockerfile for flower and entrypoint (like all doing):

version: "3.2"

services:
docker-compose.yml
  database:
    image: redis:6.2-alpine
    restart: always
    ports:
      - "6379:6379"

  rabbit:
    image: rabbitmq:3.9-management
    env_file:
      - environments/broker.env
    ports:
      - "5672:5672"
      - "15672:15672"
    restart: always

  flower:
    image: mher/flower:latest
    env_file:
      - environments/broker.env
    environment:
      - FLOWER_BASIC_AUTH=riven:an0th3r_p@ss
      - CELERY_BROKER_URL=amqp://riven:s3cr3t_p4ss@rabbit:5672/
      - FLOWER_BROKER_API_URL=http://riven:s3cr3t_p4ss@rabbit:15672/api
      - FLOWER_PORT=5555
    ports:
      - "5555:5555"
    depends_on:
      - rabbit

My flower rq.txt:

flower==1.0.0
humanize==3.13.1
prometheus-client==0.12.0
tornado==6.1

To Reproduce You need to use MACOS big sur. I guess it can make sense..

Steps to reproduce the behavior:

1. Start docker-compose and Celery (celery -A celery_test worker --loglevel=info)
2. Go to flower url and use /broker section
3. Open flower docker logs and see error.

Expected behavior Connection to RabbitMQ? WORKING?

[rive-n]

No, problem is not in CELERY_BROKER_URL=amqp://riven:s3cr3t_p4ss@rabbit:5672/. I tried to change it and give user permissions (administrator). but nothing helps. I even tried to do like in your official docs: CELERY_BROKER_URL=amqp://riven:s3cr3t_p4ss@rabbit:5672// and this trace is there too! :)

[rive-n]

I tried even to switch versions of Flower and RabbitMQ. Flower to 0.9.5 and 0.9.7 and Rabbit from alpine to manage. And nothing.

[rive-n]

and sometimes i am able to create task. But... Unknown worker '[email protected]'

[rive-n]

So if I am starting Rabbit first - i've got error about header. If second - timeout.

I am installing curl with apk get inside container. And what i am gettin?

curl -i -u riven:s3cr3t_p4ss http://rabbit:15672/api/vhosts ->

[{"cluster_state":{"rabbit@539369e1efa3":"running"},"description":"Default virtual host","message_stats":{"confirm":0,"confirm_details":{"rate":0.0},"drop_unroutable":0,"drop_unroutable_details":{"rate":0.0},"publish":1,"publish_details":{"rate":0.0},"return_unroutable":0,"return_unroutable_details":{"rate":0.0}},"messages":1,"messages_details":{"rate":0.0},"messages_ready":1,"messages_ready_details":{"rate":0.0},"messages_unacknowledged":0,"messages_unacknowledged_details":{"rate":0.0},"metadata":{"description":"Default virtual host","tags":[]},"name":"/","recv_oct":6462,"recv_oct_details":{"rate":0.0},"send_oct":8286,"send_oct_details":{"rate":0.0},"tags":[],"tracing":false}]

All works fine with it, but not with /api endpoint. It gives 301, and if i follow - just /api information.

[egelja]

Duplicate of #1112?

[rive-n]

Duplicate of #1112?

Nope, does not helps.

Errors are random and there is no solution.

And if you are talking about open duplicate - maybe you are right. And if RMQ is not working with flower - developers should write in doc about that. So i will wait for them in this thread

[j0hn-mc-clane]

Hello, Flower and Celery are both working fine with RabbitMQ, please see my example below: https://github.com/JCardoen/fast-api-example

[Comprise]

Hello, Flower and Celery are both working fine with RabbitMQ, please see my example below: https://github.com/JCardoen/fast-api-example

yes, with version 5.1.2 thanks

[rive-n]

Hello, Flower and Celery are both working fine with RabbitMQ, please see my example below: https://github.com/JCardoen/fast-api-example

Please try to reproduce my issue. Or just say what is wrong in this configuration. I can't see a real difference.

[rive-n]

Здравствуйте, Цветок и сельдерей отлично работают с RabbitMQ, пожалуйста, смотрите мой пример ниже: https://github.com/JCardoen/fast-api-example

yes, with version 5.1.2 thanks

Wdm v5.1.2?

I need to add celery to flower container with version 5.1.2?

[Comprise]

Hello, Flower and Celery are both working fine with RabbitMQ, please see my example below: https://github.com/JCardoen/fast-api-example

yes, with version 5.1.2 thanks

Wdm v5.1.2?

I need to add celery to flower container with version 5.1.2?

celery 5.1.2 + flower 1.0.0

[j0hn-mc-clane]

If this works, please close this issue.

[mira-miracoli]

I have the same issue, the broker API is not working, using flower 1.2.0 and celery 5.2.7 from my celeryconfig.py: broker_api = "https://user:pw@<broker-domain.com>/api/" Celery shows nothing in the broker tab. Otherwise its working quite fine.

[mira-miracoli]

this is the broker tab:

[mpalumbo7]

I had the same problem with the broker tab being empty. My solution was to change:

FLOWER_BROKER_API_URL=http://username:password@rabbit:15672/api

to

FLOWER_BROKER_API_URL=http://username:password@rabbit:15672/api/

Note the trailing slash.

I first came across this (annoying) requirement with Django. I'm not a fan.

[grssnbchr]

In case anyone is looking for a working config in docker-compose, here's one with RabbitMQ 3.12 and Flower 2.0.0:

  redis:
    image: redis:7.2.4
    restart: always
    networks:
      - backnet
  rabbitmq:
    image: rabbitmq:3.12-management
    ports:
      - 5672:5672
      - 15672:15672
    networks:
      - backnet
  flower:
    image: mher/flower:2.0.0
    environment:
      - CELERY_BROKER_URL=amqp://guest:guest@rabbitmq:5672//
      - CELERY_RESULT_BACKEND=redis://redis:6379/0
      - FLOWER_BROKER_API=http://guest:guest@rabbitmq:15672/api/
      - FLOWER_PORT=8888
    networks:
      - backnet
    ports:
      - 8888:8888
    depends_on:
      - rabbitmq
      - redis
networks:
  backnet:
  frontnet:

What made it work was changing the FLOWER_BROKER_API_URL into FLOWER_BROKER_API... :roll_eyes: Using Celery 5.3.6