What are the limits of ApsaraMQ for RabbitMQ - ApsaraMQ for RabbitMQ

ApsaraMQ for RabbitMQ enforces limits on clusters, characters, and peak Transactions Per Second (TPS) traffic. To avoid program errors, do not exceed these limits when using ApsaraMQ for RabbitMQ.

Cluster limits

The limits in this table apply to the latest version of instances.
If the limits of your instance do not match those in this table, your instance may not be the latest version. For more information, submit a ticket. The following restrictions apply to limit adjustments:
- Legacy architecture instances do not support limit adjustments: Instances whose ID consists of digits only, or whose VPC endpoint does not end with mq.amqp.aliyuncs.com, do not support adjustments to the limits listed below.

Item	Serverless instance		Subscription instance
	Shared	Exclusive	Enterprise Edition	Enterprise Platinum Edition	Professional Edition
	Pay-by-provisioned-capacity-and-elastic-traffic/Pay-by-messaging-request	Pay-by-provisioned-capacity-and-elastic-traffic	Enterprise Edition	Enterprise Platinum Edition	Professional Edition
Vhosts per instance	200	2,000	200	4,096	200
Exchanges per instance	6,000	10,000	6,000	80,000	6,000
Bindings per instance	20,000	20,000	20,000	None	20,000
Queues per instance	6,000	10,000	6,000	80,000	6,000
Connections per instance	10,000	20,000	Purchase based on instance type. Each instance has a free quota of 2,000 connections. The maximum is 10,000.	Purchase based on instance type. Each instance has a free quota of 50,000 connections. The maximum is 100,000.	Number of queues × 4. The number of connections per instance ranges from 1,000 to 5,000.
Queues per exchange binding	256	256 Submit a ticket to request a custom quota. After customization, you cannot downgrade the instance to a shared cluster.	256	256 To request a custom quota, contact Alibaba Cloud technical support.	256
Consumers per queue	Number of queues × 4. The number of consumers per queue ranges from 800 to 5,000.	None	Number of queues × 4. The number of consumers per queue ranges from 800 to 5,000.	None	Number of queues × 4. The number of consumers per queue ranges from 50 to 5,000.
Channels per connection	2,000	3,000	The number of channels per connection is the minimum of [Number of queues, 2048]. If the number of queues is less than 2,048, the channel limit is the number of queues. If the number of queues is 2,048 or more, the channel limit is 2,048.	3,000	The number of channels per connection is either 64 or the Number of queues. If the Number of queues is 64 or less, the number of channels per connection is 64. If the Number of queues is greater than 64, the number of channels per connection is the number of queues.
Message size	10 MB	50 MB	10 MB (Note: For instances in the China (Qingdao), Japan (Tokyo), and China (Hohhot) regions, and instances created before May 1, 2025 in other regions, the limit is 1 MB.)	50 MB (Note: For instances in the China (Qingdao), Japan (Tokyo), and China (Hohhot) regions, and instances created before May 1, 2025 in other regions, the limit is 3 MB.)	64 KB
Maximum delay for delayed messages	7 days	7 days Submit a ticket to request a custom quota. After customization, you cannot downgrade the instance to a shared cluster.	7 days	7 days To request customization, contact Alibaba Cloud Technical Support.	24 hours
Maximum message retention period	3 days	3 days	3 days	3 days	3 days
Maximum delivery attempts	16	16	16	64	16
Consumption timeout	Maximum: 3 hours Default: 5 minutes	Maximum: 12 hours Default: 30 minutes	Maximum: 3 hours Default: 5 minutes	Maximum: 12 hours Default: 30 minutes	Maximum: 10 minutes Default: 1 minute

Note

ApsaraMQ for RabbitMQ limits the size of a message Header frame to 32 KB. This limit cannot be adjusted. To avoid triggering the Header frame size limit, include large message content in the message Body rather than the Header frame.

Character limits

Item	Limit
Queue name	Can contain only letters, digits, hyphens (-), underscores (_), periods (.), number signs (#), forward slashes (/), and at signs (@). Must be 1 to 255 characters in length.
Exchange name	Can contain only letters, digits, hyphens (-), underscores (_), periods (.), number signs (#), forward slashes (/), and at signs (@). Must be 1 to 255 characters in length.
Vhost name	Can contain only letters, digits, hyphens (-), underscores (_), periods (.), number signs (#), forward slashes (/), and at signs (@). Must be 1 to 255 characters in length.
Binding key	Standard type Can contain only letters, digits, hyphens (-), underscores (_), periods (.), forward slashes (/), and at signs (@). Must be 1 to 255 characters in length. Topic type Can contain only letters, digits, hyphens (-), underscores (_), asterisks (), periods (.), number signs (#), forward slashes (/), and at signs (@). Must be 1 to 255 characters in length. Cannot start or end with a period (.). If the key starts with a number sign (#) or an asterisk (), it must be followed by a period. If the key ends with one of these characters, it must be preceded by a period. If one of these characters is in the middle of the key, it must be enclosed by periods.
Consumer tag	Can contain only letters, digits, hyphens (-), underscores (_), periods (.), number signs (#), forward slashes (/), and at signs (@). Must be 1 to 255 characters in length.
Message ID	Can contain only letters, digits, hyphens (-), underscores (_), periods (.), number signs (#), forward slashes (/), at signs (@), and colons (:). Must be 1 to 255 characters in length.

Rate limiting thresholds

Important

For information about how TPS is calculated, see Calculation rules for Serverless instance editions and TPS calculation rules for subscription instance editions.
For information about how to resolve issues when TPS is rate-limited, see Solutions for when the total TPS of an instance is rate-limited and Solutions for when the TPS of a single node is rate-limited.

Total TPS

Instance edition	Serverless instance		Subscription instance
Specification	Shared	Exclusive	Elastic TPS disabled			Elastic TPS enabled
Specification	Pay-by-provisioned-capacity-and-elastic-traffic/Pay-by-messaging-request	Pay-by-provisioned-capacity-and-elastic-traffic	Enterprise Edition	Enterprise Platinum Edition	Professional Edition	Enterprise Edition	Enterprise Platinum Edition	Professional Edition
Rate limiting threshold	Maximum 50,000 per second	Twice the peak TPS of the base specifications	Peak TPS of the base specifications			Twice the peak TPS of the base specifications, up to a maximum of 50,000 per second	Twice the peak TPS of the base specifications, up to a maximum of 50,000 per second	1.5 times the peak TPS of the base specifications

Single-node SendMessage

The server limits the SendMessage TPS for each backend service node in an instance. The thresholds are as follows:

Item	Serverless instance			Subscription instance
	Shared		Exclusive	Enterprise Edition	Enterprise Platinum Edition	Professional Edition
	Pay-by-messaging-request	Pay-by-provisioned-capacity-and-elastic-traffic	Pay-by-provisioned-capacity-and-elastic-traffic
Rate limiting threshold	25,000 per second	25,000 per second	None	25,000 per second	None	25,000 per second

Single API call

Item	API	Serverless instance		Subscription instance
		Shared	Exclusive	Enterprise Edition	Enterprise Platinum Edition	Professional Edition
		Pay-by-provisioned-capacity-and-elastic-traffic/Pay-by-messaging-request	Pay-by-provisioned-capacity-and-elastic-traffic	Enterprise Edition	Enterprise Platinum Edition	Professional Edition
Purge queue per instance	`purgeQueue`	500 TPS	None	500 TPS	None	500 TPS
Create exchange per instance	`exchangeDeclare`	500 TPS	None	500 TPS	None	500 TPS
Delete exchange per instance	`exchangeDelete`	500 TPS	None	500 TPS	None	500 TPS
Create queue per instance	`queueDeclare`	500 TPS	None	500 TPS	None	500 TPS
Delete queue per instance	`queueDelete`	500 TPS	None	500 TPS	None	500 TPS
Create binding per instance	`queueBind`	500 TPS	None	500 TPS	None	500 TPS
Delete binding per instance	`queueUnbind`	500 TPS	None	500 TPS	None	500 TPS
Single-Instance Recovery Message	`basicRecover`	500 TPS	None	500 TPS	None	500 TPS
Requeue messages per instance	`basicReject(requeue=true)` `basicNack(requeue=true)`	20 TPS	None	20 TPS	None	20 TPS

FAQ

Does Enterprise Edition support independently expanding the queue limit?

No. The maximum number of queues per Enterprise Edition instance is 6,000. This limit is a product-tier specification and cannot be independently expanded within the current edition (for example, it cannot be independently raised to 10,000).

If your business requires more queues, consider the following alternatives:

Optimize queue usage: Consolidate queues where possible, use Topic Exchange routing strategies to reduce the total number of queues, or clean up queues that have no consumers and are no longer in use.
Split the business across instances: If your workload allows, distribute part of your business to a new Enterprise Edition instance to spread queue usage across multiple instances.

How do I handle the error "The channelMax limit is reached"?

Cause: The number of channels under a single connection has reached the channelMax limit for your current instance edition.

Solution:

Confirm your current ApsaraMQ for RabbitMQ instance edition (Professional Edition, Enterprise Edition, Enterprise Platinum Edition, and so on).
Refer to the Channels per connection row in the Cluster limits table to find the specific channelMax value for your instance edition.
Evaluate whether you need to upgrade your instance edition to support a higher number of concurrent channels.

Note

The keywords "channelMax", "limit is reached", "pay callback error", and "channel count limit" all relate to this error. If you encounter any of these symptoms, follow the steps above to diagnose and resolve the issue.