Task #1370

Updated by rbarlow almost 6 years ago

Pulp does support RabbitMQ, but we historically have not no QE'd Pulp with RabbitMQ and our documentation guides users towards using qpidd. Celery recommends that their users use RabbitMQ, and most upstream Celery deployments use it. Due to this, I think we can have high confidence that Pulp will work well with RabbitMQ though it is untested.

This task is primarily to ensure that Pulp gets tested with RabbitMQ, with a few potential secondary objectives:
* Develop a plan to automatically QE Pulp with RabbitMQ (in addition to qpidd)
* Rework our documentation to treat RabbitMQ at least equitably with qpidd, but leave qpidd as the default in server.conf
* Give users a nice table to compare RabbitMQ fairly against qpidd with all the relevant pros
and cons:
- qpidd has historically been much more widely used with Pulp, in development, upstream, and downstream (can we mention downstream in the docs?)
consider recommending RabbitMQ is widely used with Celery
- RabbitMQ also supports vhosts, which is an easier way
to separate workers from consumers
* Look into run pulp-smash against Fedora's updates-testing repos
users as well, to catch broker issues before they are released so that we can report karma when necessary a default (since Celery makes this recommendation)
* If all goes well, consider recommending the use of RabbitMQ to the Fedora Project infrastructure.