What’s new?

The naming convention for DLT topics has been standardized to use the "-dlt" suffix consistently. This change ensures compatibility and avoids conflicts when transitioning between different retry solutions. Users who wish to retain the ".DLT" suffix behavior need to opt-in explicitly by setting the appropriate DLT name property.

A new method, getGroupId(), has been added to the ConsumerSeekCallback interface. This method allows for more selective seek operations by targeting only the desired consumer group. The AbstractConsumerSeekAware can also now register, retrieve, and remove all callbacks for each topic partition in a multi-group listener scenario without missing any. See the new APIs (getSeekCallbacksFor(TopicPartition topicPartition), getTopicsAndCallbacks()) for more details. For more details, see Seek API Docs.

RecordFilterStrategy now supports ignoring empty batches that result from filtering. This can be configured through overriding default method ignoreEmptyBatch(), which defaults to false, ensuring KafkaListener is invoked even if all ConsumerRecords are filtered out. For more details, see Message receive filtering Docs.

The ConcurentContainerMessageListenerContainer emits now a ConcurrentContainerStoppedEvent when all of its child containers are stopped. For more details, see Application Events and ConcurrentContainerStoppedEvent Javadocs.

When using ReplyingKafkaTemplate, if the original record from the request contains a key, then that same key will be part of the reply as well. For more details, see Sending Messages section of the reference docs.

When using DeadLetterPublishingRecovererFactory, the user applications can override the maybeLogListenerException method to customize the logging behavior.

When extending KafkaAdmin, user applications may override the createAdmin method to customize Admin client creation.

When using KafkaStreamsCustomizer it is now possible to return a custom implementation of the KafkaStreams object by overriding the initKafkaStreams method.

When using a BatchListener, the ConsumerRecord can have the KafkaHeaders.DELIVERY_ATTMPT header in its headers fields. If the DeliveryAttemptAwareRetryListener is set to error handler as retry listener, each ConsumerRecord has delivery attempt header. For more details, see Kafka Headers for Batch Listener.

The MicrometerProducerListener, MicrometerConsumerListener and KafkaStreamsMicrometerListener can now be configured with a TaskScheduler. See KafkaMetricsSupport JavaDocs and Micrometer Support for more information.