RabbitMQ: How to move failed message from one queue to another queue?

Question

I have two queues:

Same is visible when I run rabbitmqadmin list queues vhost name node messages message_stats.publish_details.rate -u admin -p admin:

I get:

+-------+-------------------------+-------------------------+----------+------------------------------------+
| vhost |          name           |          node           | messages | message_stats.publish_details.rate |
+-------+-------------------------+-------------------------+----------+------------------------------------+
| /     | high_priority           | rabbit@server-rabbitmq  | 5        | 0.0                                |
| /     | high_priority_secondary | rabbit@server-rabbitmq  | 0        | 0.0                                |
+-------+-------------------------+-------------------------+----------+------------------------------------+

My exchanges (rabbitmqadmin -V / list exchanges -u admin -p admin) are listed below:

+-------------------------+---------+
|          name           |  type   |
+-------------------------+---------+
|                         | direct  |
| amq.direct              | direct  |
| amq.fanout              | fanout  |
| amq.headers             | headers |
| amq.match               | headers |
| amq.rabbitmq.trace      | topic   |
| amq.topic               | topic   |
| high_priority           | direct  |
| high_priority_secondary | direct  |
| low_priority            | direct  |
+-------------------------+---------+

Queues and the whole related logic is implemented in PHP / Symfony, however I would like to use native logic (if possible) by using either rabbitmqadmin or rabbitmqctl commands in terminal.

If a message on the high_priority fails, I would like RabbitMQ to automatically move it to the high_priority_secondary queue without any PHP involvement. Is this possible? I've started reading about Dead Letter Exchanges but I'm not sure how to approach this.

I already created a consumer for the secondary queue so as soon as the message is moved there, it will be processed.

Is is possible to achieve this in CLI only?

FYI: There are some suggested posts on SO that already cover the question but none of the solutions is purely CLI one.

Answer 1

OK, while I didn't have to modify any PHP code, I did have to alter yaml configuration on the framework level as I wanted my solution to be persisted and part of the code base.

In your app/config/services/rabbitmq.yaml:

Define producers:

high_priority:
    connection: default
    class: Foo\Infrastructure\RabbitMQ\SuppressedProducer
    exchange_options:
        name: 'high_priority'
        type: direct
high_priority_secondary:
    connection: default
    class: Foo\Infrastructure\RabbitMQ\SuppressedProducer
    exchange_options:
        name: 'high_priority_secondary'
        type: direct
message_hospital:
    connection: default
    class: Foo\Infrastructure\RabbitMQ\SuppressedProducer
    exchange_options:
        name: 'message_hospital'
        type: direct

Define consumers:

high_priority:
    connection: default
    exchange_options:
        name: 'high_priority'
        type: direct
    queue_options:
        name: 'high_priority'
        arguments:
            x-dead-letter-exchange: ['S', 'high_priority_secondary']
    qos_options:
        prefetch_size: 0
        prefetch_count: 1
        global: false
    callback: foo.task_bus.consumer
high_priority_secondary:
    connection: default
    exchange_options:
        name: 'high_priority_secondary'
        type: direct
    queue_options:
        name: 'high_priority_secondary'
        arguments:
            x-dead-letter-exchange: ['S', 'message_hospital']
    qos_options:
        prefetch_size: 0
        prefetch_count: 1
        global: false
    callback: foo.task_bus.consumer
message_hospital:
    connection: default
    exchange_options:
        name: 'message_hospital'
        type: direct
    queue_options:
        name: 'message_hospital'
    qos_options:
        prefetch_size: 0
        prefetch_count: 1
        global: false
    callback: foo.task_bus.consumer

Now the queues look like:

Thanks to the DLX attributes the message goes to the hospital queue as soon as it fails in the previous ones.