Amazon MQ for RabbitMQ broker instance types

The combined description of the broker instance class (m7g) and size (large, medium) is called the broker instance type (for example, mq.m7g.large).

We recommend using mq.m7g instance types for both cluster and single-instance deployments.

Amazon MQ provides at least a 90 day notice before an instance type reaches end of support. We recommend upgrading your broker to a new instance type before the end-of-support date to prevent any disruptions.

Instance types for m7g cluster deployment

We recommending using mq.m7g.x instance types with cluster deployment. The following table shows the available mq.m7g.x instance types for cluster deployment.

Instance types for m7g single instance deployment

The following table shows the available mq.m7g.x instance types for single instance deployment.

Instance types for mq.m5 single instance deployment

The following tables show the available mq.m5.x instance types for single instance deployment

Instance types for mq.m5 cluster deployment

The following tables show the available mq.m5.x instance types for cluster deployment

Memory and disk alarms

Amazon MQ configures memory and disk thresholds on each RabbitMQ broker to protect against resource exhaustion. When a threshold is exceeded, RabbitMQ triggers an alarm and blocks publishers from sending messages. Consumers on separate connections continue to operate normally. However, if a publisher and consumer share the same connection, the consumer is also blocked.

Memory alarm

The vm_memory_high_watermark parameter defines the maximum amount of memory that a RabbitMQ broker can use before it blocks publishers from sending messages. When memory usage exceeds this threshold, RabbitMQ triggers a memory alarm. For more information, see Memory Alarms on the RabbitMQ website.

For mq.m7g instance types, Amazon MQ sets the following absolute memory high watermark values:

For mq.m5 instance types, Amazon MQ sets a relative memory high watermark of 0.4 (40% of the available memory).

The higher memory thresholds on mq.m7g instances allow RabbitMQ to use more available memory before triggering an alarm. For more information about performance improvements with mq.m7g instances, see Improve RabbitMQ performance on Amazon MQ with AWS Graviton3-based M7g instances on the AWS Blog.

Disk alarm

The disk_free_limit parameter defines the minimum amount of free disk space that a RabbitMQ node requires. When free disk space on any node drops below this limit, RabbitMQ triggers a disk alarm and blocks publishers from sending messages. For more information, see Disk Alarms on the RabbitMQ website.

For mq.m7g instance types, Amazon MQ sets the following disk free limits. Single-instance brokers have a higher disk free limit to provide additional protection because they do not have other nodes to serve traffic if disk space is exhausted.

For mq.m5 instance types, Amazon MQ sets the following disk free limits. These values apply to both single-instance and cluster deployments.

Because mq.m7g instances have a lower disk free limit, more of the provisioned disk volume is available for message storage compared to equivalent mq.m5 instances.