The purpose of this PR is to enable concurrent handling of Symfony Messenger messages. This PR utilizes the amphp/parallel library to achieve this. During the worker configuration, it is possible to define how many processes should be batched in parallel before flushing, meaning accessing the result, acknowledging, or initiating executions in case of errors.

In the child process, a container is cached. Therefore, in each child process, it will be possible to inject services and make requests, etc.

In this initial use case, the decision was made not to reuse the parent connection, such as the Doctrine connection. This is because it could currently be inconvenient for users, as they would need to modify their handlers, which could be cumbersome.

Even without reusing the parent connection, there is a performance gain since operations are performed concurrently for x number of processes. Therefore, even if there is a handler with blocking code, it does not prevent other child processes from proceeding.
for example, in the case we have 2 handlers and one of the handlers has a 4-second pause, and we have, for instance, 40 messages processed concurrently with a batch size of 10 using the ParallelBus, it's approximately 6 times faster between handling the first message and the last message (This observation was made during a test conducted in our development environment)

In order to dispatch messages concurrently, it is necessary to first consider the ParallelMessageBus:
! It works for async mode - It doesn't work with BatchHandler Trait !

Async mode:

$this->bus->dispatch(newUserMessage($user->getId()), [newBusNameStamp('parallel_bus')]);

You can specify how many processes will be executed in parallel.
It will define how many processes will be launched - threaded to then complete them one by one and get the return of the message processing
By default: 10

php bin/console messenger:consume async -p 20

TODO:

• CHANGELOG:

This feature is being developed by@coopTilleuls and@TradersPost and it has been designed with@jwage and@dunglas