AMQP Channels
Channels are lightweight virtual connections multiplexed over a single AMQP connection. All AMQP operations (publish, consume, declare, etc.) happen on a channel.
Channels are an AMQP concept and do not apply to MQTT.
Why Channels Exist
Opening a TCP connection is expensive. Channels allow multiple independent streams of communication over one connection, avoiding the overhead of multiple TCP connections while maintaining isolation between operations.
Channel Limits
The maximum number of channels per connection is negotiated during connection setup; the lower of the client and server values wins.
| Config Key | Section | Default | Description |
|---|---|---|---|
channel_max |
[amqp] |
2048 |
Maximum channels per connection |
Prefetch (QoS)
Prefetch controls how many unacknowledged messages the server will deliver to a consumer before waiting for acknowledgments. Set via basic.qos.
| Parameter | Description |
|---|---|
prefetch_count |
Max unacknowledged messages. 0 means unlimited. |
global |
If false (default): per-consumer prefetch. If true: per-channel prefetch shared across all consumers on the channel. |
The server-wide default applies when a consumer does not call basic.qos:
| Config Key | Section | Default | Description |
|---|---|---|---|
default_consumer_prefetch |
[main] |
65535 |
Default per-consumer prefetch |
Global Prefetch
With global set to true, the limit counts unacknowledged deliveries across all consumers on the channel. It applies on top of each consumer’s own prefetch, so a delivery needs room under both limits. Messages fetched with basic.get don’t count towards it, and a prefetch_count of 0 removes the channel limit.
When the channel is at its limit, its consumers wait in line. Each delivery that is acknowledged or rejected frees a slot for the consumer that has waited longest, so deliveries are spread across the queues on the channel instead of one queue being drained first. This keeps a consumer from being starved; it isn’t a guarantee of an even split.
Lowering the limit below the current number of unacknowledged deliveries stops new deliveries until enough of them have been acknowledged.
Channel Errors
When a channel error occurs (e.g., accessing a non-existent queue, argument mismatch), the server closes the channel with an error code and message. The connection remains open and other channels are unaffected.
Common channel-level errors:
- NOT_FOUND — resource does not exist
- PRECONDITION_FAILED — argument mismatch on redeclaration
- ACCESS_REFUSED — permission denied
- RESOURCE_LOCKED — exclusive queue owned by another connection
Flow Control
Clients can pause message delivery on a channel by sending channel.flow with active=false. The server responds with channel.flow-ok. When re-enabled (active=true), delivery resumes.
When the server runs low on disk space, basic.publish returns a precondition_failed channel error until resources recover.
Consumer Limit
| Config Key | Section | Default | Description |
|---|---|---|---|
max_consumers_per_channel |
[amqp] |
0 |
Max consumers per channel. 0 means unlimited. |
Ready to take the next steps?
Managed LavinMQ instance via CloudAMQP
LavinMQ has been built with performance and ease of use in mind - we've benchmarked a throughput of about 1,000,000 messages/sec. You can try LavinMQ without any installation hassle by creating a free instance on CloudAMQP. Signing up is a breeze.
Get started with CloudAMQP ->Help and feedback
We welcome your feedback and are eager to address any questions you may have about this piece or using LavinMQ. Join our Slack channel to connect with us directly. You can also find LavinMQ on GitHub.