Question about Spring AMQP, DLQ, and Error Handling

[mimparatta]

Hi,

I’m working on a personal project with two Spring Boot applications communicating asynchronously via RabbitMQ using Spring AMQP.

I successfully set up communication and explored error handling based on this article: https://www.baeldung.com/spring-amqp-error-handling. After implementing the mechanism and testing, I noticed this behavior:

When a fatal exception occurs (determined by overriding isFatal), the message is sent to the DLQ, processed there, then returned to the original queue before being discarded. Is this expected? Shouldn’t it be discarded immediately instead?

After debugging, I found that in ConditionalRejectingErrorHandler.handleError (in spring-rabbit), the x-death header is null when the message fails initially but is set when it returns from the DLQ, after which it gets discarded.

Could someone clarify if this is the expected behavior?

Thanks!

[artembilan]

The behavior is correct.
See ConditionalRejectingErrorHandler Javadocs and its logic:

	/**
	 * Set to false to disable the (now) default behavior of logging and discarding
	 * messages that cause fatal exceptions and have an `x-death` header; which
	 * usually means that the message has been republished after previously being
	 * sent to a DLQ.
	 * @param discardFatalsWithXDeath false to disable.
	 * @since 2.1
	 */
	public void setDiscardFatalsWithXDeath(boolean discardFatalsWithXDeath) {

So, the default is exactly what you have observed: the message is sent to DLX, then it is comes back to the original queue, but it has now that x-death header.
Therefore, when this message fails again, we discard it.

You may treat this logic as a retry pattern.
We give the failed message a chance to come back for possible normal processing.

[mimparatta]

hi @artembilan

I understand.
thanks for the explanation.