How to implement symfony messenger

[tdumalin]

Hi,

I've been looking on this project issues and I found some answers but I think it would be a good Idea to document properly how to use Symfony Messenger using this template.
Syfmony Messenger is a key component (imo).
It's already used for very useful async feature such as:

• Email sending
• Symfony scheduler (cool cron replacement)

Here's my setup:

# compose.prod.yml:
version: "3.4"

# Production environment override
services:
   php: #...
     
    messenger:
        build:
            context: .
            target: frankenphp_prod_messenger
        command: 'sh -c "php bin/console messenger:consume async scheduler_default -vv"'
        environment:
            DATABASE_URL: xxxx
            MAILER_DSN: xxxx
        restart: unless-stopped
        depends_on:
            - php     

# Dockerfile

# all existing stuff ...

## Messenger
FROM frankenphp_prod AS frankenphp_prod_messenger
WORKDIR /app

HEALTHCHECK NONE

ENTRYPOINT ["docker-entrypoint"]

# config/packages/messenger.yaml
framework:
    messenger:
        failure_transport: failed

        transports:
            # https://symfony.com/doc/current/messenger.html#transport-configuration
            async: '%env(MESSENGER_TRANSPORT_DSN)%'
            failed: 'doctrine://default?queue_name=failed'
            sync: 'sync://'

        routing:
            Symfony\Component\Mailer\Messenger\SendEmailMessage:  async
            # other messages ...

#in dev env, we use sync transport to process messages immediately
when@dev:
    framework:
        messenger:
            transports:
                async: 'sync://'

But according to the Symfony documentation, worker should not run forever and should be managed by supervisor or systemd

What do you guys think ? How should we implements messenger in this template following the Symfony doc advices ?

Thanks for the help

[MichaelBrauner]

Hi tdumalin,

You might consider using the --limit option with messenger:consume:

php-worker:
  build:
    context: .
    target: frankenphp_dev
  image: ${IMAGES_PREFIX:-}app-php
  restart: unless-stopped
  command: php bin/console messenger:consume async -v --limit=10000

With this setup, the restart: unless-stopped Docker directive ensures that the container is automatically restarted after processing the specified number of messages.

This approach avoids the need for external process control systems.
Please let me know what you think about it and how you solved this issue (which has been a while now).

[tdumalin]

Hi @MichaelBrauner ,

Sounds great, works like a charm.
I think it should definitely be documented somewhere

Thanks for the help

[leroy0211]

If you are running this in other orchestrator like kubernetes, you might get the effect that your container/pod doesn't start until a period of time has expired.
Kubernetes considers a pod that stops as a "crashed" pod. Every time a pod "crashes" an increment of a few seconds is used to restart your pod. This could end up in kubernetes thinking your pod is faulty and doesn't start your pod for up to minutes!
Kubernetes uses a reset mechanism. So if you pod runs for 10 minutes straight, the "crashed" timer resets.

When you're using kubernets you need something like supervisord, a simple while true loop in the main process, or increment the settings of the messenger:consume command so that "crashed" stat is always reset after a restart, to prevent your pod from ending up in a "CrashedLoopback" state.

I haven't found any configuration in a kuberneted pod definition to stop this, or lower the reset timer to just a few seconds.