Stressing RabbitMQ

We have classified in Chapter 1, Designing OpenStack Cloud Architecture, the database and queuing message system as very critical components in the OpenStack environment. If you have already found different ways to boost your database and ensured that is performing well, you will need on the other hand to measure the RabbitMQ capacity so you can identify any bottleneck at an early stage. Although we have clustered our message queuing system, we should take into account that if one of the nodes in the cluster becomes down or unreachable, the remaining one can face a sudden heavy workload which may lead to a bottleneck. Then what? All the OpenStack services will not be able to talk to RabbitMQ and the entire cluster stops working. Basically, when adding new compute nodes simultaneously to the controller node, RabbitMQ will need to create more processes and threads to be able to manage the new compute services and join them to talk to other running OpenStack services. Default...