RabbitMQ 4.0 released with Khepri support, enhanced AMQP 1.0, and improved queue features
RabbitMQ has launched its latest major release, RabbitMQ 4.0, introducing several significant updates and improvements. One of the key changes is the full support for Khepri, a schema data store developed to replace Mnesia, which was previously an experimental feature.
AMQP 1.0 has become a core protocol and is now always enabled, boasting a peak throughput more than double that of version 3.13.x for certain workloads. AMQP 1.0 clients can now manage topologies similarly to AMQP 0-9-1 clients, with an improved address format for easier interaction with AMQP 0-9-1 entities.
RabbitMQ 4.0 also features an efficient sub-linear quorum queue recovery on node startup using checkpoints and introduces priority support for quorum queues, although not in the same manner as classic queues. The mirroring of classic queues has been removed, pushing users towards quorum queues and streams for replicated messaging data types. However, non-replicated classic queues remain and continue to be developed.
The new version also brings enhancements in classic queue storage efficiency, notably in recovery time and the handling of multi-MiB messages. Nodes with multiple enabled plugins and minimal on-disk data now experience a 20-30% faster startup. Additionally, a new exchange type, the Local Random Exchange, has been introduced.