Amazon MQ for RabbitMQ broker defaults - Amazon MQ
Policy and limit descriptionsRecommended default values

Amazon MQ for RabbitMQ broker defaults

When you create an Amazon MQ for RabbitMQ broker, Amazon MQ applies a default set of broker policies and vhost limits to optimize your broker's performance. Amazon MQ applies vhost limits only to the default (/) vhost. Amazon MQ will not apply default policies to newly created vhosts. We recommend keeping these defaults for all new and existing brokers. However, you can modify, override, or delete these defaults at any time.

Amazon MQ creates policies and limits based on the instance type and broker deployment mode that you choose when you create your broker. The default policies are named according to the deployment mode, as follows:

  • Single-instance – AWS-DEFAULT-POLICY-SINGLE-INSTANCE

  • Cluster deployment – AWS-DEFAULT-POLICY-CLUSTER-MULTI-AZ

For single-instance brokers, Amazon MQ sets the policy priority value to 0. To override the default priority value, you can create your own custom policies with higher priority values. For cluster deployments, Amazon MQ sets the priority value to 1 for broker defaults. To create your own custom policy for clusters, assign a priority value greater than 1.

Note

In cluster deployments, ha-mode and ha-sync-mode broker policies are required for classic mirroring and high availability (HA).

If you delete the default AWS-DEFAULT-POLICY-CLUSTER-MULTI-AZ policy, Amazon MQ uses the ha-all-AWS-OWNED-DO-NOT-DELETE policy with a priority value of 0. This ensures that the required ha-mode and ha-sync-mode policies are still in effect. If you create your own custom policy, Amazon MQ automatically appends ha-mode and ha-sync-mode to your policy definitions.

Policy and limit descriptions

The following list describes the default policies and limits that Amazon MQ applies to a newly created broker. The values for max-length, max-queues, and max-connections vary based on your broker's instance type and deployment mode. These values are listed in the Recommended default values section.

  • queue-mode: lazy (policy) – Enables lazy queues. By default, queues keep an in-memory cache of messages, enabling the broker to deliver messages to consumers as fast as possible. This can lead to the broker running out of memory and raising a high-memory alarm. Lazy queues attempt to move messages to disk as early as is practical. This means that fewer messages are kept in memory under normal operating conditions. Using lazy queues, Amazon MQ for RabbitMQ can support much larger messaging loads and longer queues. Note that for certain use cases, brokers with lazy queues might perform marginally slower. This is because messages are moved from disk to broker, as opposed to delivering messages from an in-memory cache.

    Deployment modes

    Single-instance, cluster

  • max-length: number-of-messages (policy) – Sets a limit for the number of messages in a queue. In cluster deployments, the limit prevents paused queue synchronization in cases such as broker reboots, or following a maintenance window.

    Deployment modes

    Cluster

  • overflow: reject-publish (policy) – Enforces queues with a max-length policy to reject new messages after the number of messages in the queue reaches the max-length value. To ensure that messages aren't lost if a queue is in an overflow state, client applications that publish messages to the broker must implement publisher confirms. For information about implementing publisher confirms, see Publisher Confirms on the RabbitMQ website.

    Deployment modes

    Cluster

  • max-queues: number-of-queues-per-vhost (vhost limit) – Sets the limit for the number of queues in a broker. Similar to the max-length policy definition, limiting the number of queues in cluster deployments prevents paused queue synchronization following broker reboots or maintenance windows. Limiting queues also prevents excessive amounts of CPU usage for maintaining queues.

    Deployment modes

    Single-instance, cluster

  • max-connections: number-of-connections-per-vhost (vhost limit) – Sets the limit for the number of client connections to the broker. Limiting the number of connections according to the recommended values prevents excessive broker memory usage, which could result in the broker raising a high memory alarm and pausing operations.

    Deployment modes

    Single-instance, cluster

Recommended default values

Note

The max-length and max-queue default limits are tested and evaluated based on an average message size of 5 kB. If your messages are significantly larger than 5 kB, you will need to adjust and reduce the max-length and max-queue limits.

The following table lists the default limit values for a newly created broker. Amazon MQ applies these values according to the broker's instance type and deployment mode.

Instance type Deployment mode max-length max-queues max-connections
t3.micro Single-instance N/A 500 500
m5.large Single-instance N/A 20,000 4,000
Cluster 8,000,000 4,000 15,000
m5.xlarge Single-instance N/A 30,000 8,000
Cluster 9,000,000 5,000 20,000
m5.2xlarge Single-instance N/A 60,000 15,000
Cluster 10,000,000 6,000 40,000
m5.4xlarge Single-instance N/A 150,000 30,000
Cluster 12,000,000 10,000 100,000