How to Prevent RabbitMQ Overload with Consumer QoS and Manual ACK

Problem: Message overload in RabbitMQ consumers

When a RabbitMQ server has tens of thousands of unprocessed messages, opening a consumer causes all messages to be pushed at once, overwhelming the client and potentially crashing the server. Similar imbalance appears when a producer generates hundreds of messages per minute while a single consumer can only handle about 60.

Solution: Consumer‑side flow control

RabbitMQ provides a QoS (quality‑of‑service) feature that limits the number of unacknowledged messages a consumer may receive. By disabling automatic acknowledgments (autoAck = false) and setting a prefetch count, the broker stops delivering new messages until the consumer acknowledges the previous ones.

QoS API

The relevant method is:

void BasicQos(uint prefetchSize, ushort prefetchCount, bool global);

prefetchSize

: size limit per message, usually set to 0 (no limit). prefetchCount: maximum number of messages the consumer can handle simultaneously. global: true applies the setting at the channel level; false applies it per consumer.

autoAck must be set to false; the QoS limit only works when manual acknowledgments are used.

Manual acknowledgment

After processing a message, the consumer calls:

void basicAck(Integer deliveryTag, boolean multiple);

multiple

indicates batch acknowledgment; in this tutorial it is set to false because messages are processed one by one.

Practical code walk‑through

The article shows a custom consumer that enables QoS with basicQos(0, 1, false), processes a single message, and sends a manual ACK. Screenshots illustrate the consumer and producer consoles before and after enabling QoS and ACK, demonstrating that only one message is processed at a time while the remaining messages stay pending until ACK is sent.

By applying this flow‑control mechanism, the consumer avoids being flooded, the broker stays stable, and the system can handle high‑volume production without crashes.