Sometimes messages can be filtered out before the deserialization based on header values . Are there any existing patterns for this scenario using spring kafka. I am thinking implementing similar to ErrorHandlingDeserializer in addition to delegate take filter predicate also as property. Any suggestions? thanks.
Yes, you can use the same technique used by the ErrorHandlingDeserializer to return a "marker" object instead of doing the deserialization, then add a RecordFilterStrategy, that filters records with such objects, to the listener (container factory when using #KafkaListener or use a filtering adapter for an explicit listener).
EDIT
Spring Boot and adding a filter...
#Bean
public ConcurrentKafkaListenerContainerFactory<?, ?> kafkaListenerContainerFactory(
ConcurrentKafkaListenerContainerFactoryConfigurer configurer,
ConsumerFactory<Object, Object> kafkaConsumerFactory) {
ConcurrentKafkaListenerContainerFactory<Object, Object> factory = new ConcurrentKafkaListenerContainerFactory<>();
configurer.configure(factory, kafkaConsumerFactory);
kafkaConsumerFactory.setRecordFilterStrategy(myFilter());
return factory;
}
Related
I am using spring.
I have a configured ObjectMapper for the entire project and I use it to set up a kafka deserializer.
And then I need a custom kafka deserializer to be used in KafkaListener.
I'm configuring KafkaListener via autoconfiguration, not via #Configuration class.
#Component
#RequiredArgsConstructor
public class CustomMessageDeserializer implements Deserializer<MyMessage> {
private final ObjectMapper objectMapper;
#SneakyThrows
#Override
public MyMessage deserialize(String topic, byte[] data) {
return objectMapper.readValue(data, MyMessage.class);
}
}
If i do like this
#KafkaListener(
topics = {"${topics.invite-user-topic}"},
properties = {"value.deserializer=com.service.deserializer.CustomMessageDeserializer"}
)
public void receiveInviteUserMessages(MyMessage myMessage) {}
I received KafkaException: Could not find a public no-argument constructor
But with public no-argument constructor in CustomMessageDeserializer class i am getting NPE because ObjectMapper = null. It creates and uses a new class, not a spring component.
#KafkaListener supports SpEL expressions.
And I think that this problem can be solved using SpEL.
Do you have any idea how to inject spring bean CustomMessageDeserializer with SpEL?
There are no easy ways to do it with SPeL.
Analysis
To get started, see the JavaDoc for #KafkaListener#properties:
/**
*
* SpEL expressions must resolve to a String ...
*/
The value of value.deserializer is used to instantiate the specified deserializer class. Let's follow the call chain:
You specify this value in the #KafkaListener annotation, then you are probably not creating a bean of the ConsumerFactory.class. So Spring creates this bean class itself - see KafkaAutoConfiguration#kafkaConsumerFactory.
Next is the creation of the returned object new DefaultKafkaConsumerFactory(...) as ConsumerFactory<?,?> using the constructor for default delivery expressions keyDeserializer/valueDeserializer = () -> null
This factory is used to create a Kafka consumer (The entry point is the constructor KafkaMessageListenerContainer#ListenerConsumer, then KafkaMessageListenerContainer.this.consumerFactory.createConsumer...)
In the KafkaConsumer constructor, the valueDeserializer object is being created, because it is null (for the default factory of point 2 above):
if (valueDeserializer == null) {
this.valueDeserializer = config.getConfiguredInstance(ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG, Deserializer.class);
The implementation of config.getConfiguredInstance involves instantiating your deserializer class via a parameterless constructor using reflection and your String "com.service.deserializer.CustomMessageDeserializer" class name
Solutions
To use value.deserializer with your customized ObjectMapper, you must create the ConsumerFactory bean yourself using the setValueDeserializer(...) method. This is also mentioned in the second Important part of the JSON.Mapping_Types.Important documentation
If you don't want to create a ConsumerFactory bean, and also don't have complicated logic in your deserializer (you only have return objectMapper.readValue(data, MyMessage.class);), then register DefaultKafkaConsumerFactoryCustomizer:
#Bean
// inject your custom objectMapper
public DefaultKafkaConsumerFactoryCustomizer customizeJsonDeserializer(ObjectMapper objectMapper) {
return consumerFactory ->
consumerFactory.setValueDeserializerSupplier(() ->
new org.springframework.kafka.support.serializer.JsonDeserializer<>(objectMapper));
}
In this case, you don't need to create your own CustomMessageDeserializer class (remove it) and Spring will automatically parse the message into your MyMessage.
#KafkaListener annotation should also not contains the property properties = {"value.deserializer=com.my.kafka_test.component.CustomMessageDeserializer"}. This DefaultKafkaConsumerFactoryCustomizer bean will automatically be used to configure the default ConsumerFactory<?, ?> (see the implementation of the KafkaAutoConfiguration#kafkaConsumerFactory method)
Here how it works for me:
#KafkaListener(topics = "${solr.kafka.topic}", containerFactory = "batchFactory")
public void listen(List<SolrInputDocument> docs, #Header(KafkaHeaders.BATCH_CONVERTED_HEADERS) List<Map<String, Object>> headers, Acknowledgment ack) throws IOException {...}
And then I have 2 beans defined in my Configuration
#Profile("!test")
#Bean
#Autowired
public ConsumerFactory<String, SolrInputDocument> consumerFactory(KafkaProperties properties) {
Map<String, Object> props = properties.buildConsumerProperties();
props.put(ConsumerConfig.ENABLE_AUTO_COMMIT_CONFIG, false);
DefaultKafkaConsumerFactory<String, SolrInputDocument> result = new DefaultKafkaConsumerFactory<>(props);
String validatedKeyDeserializerName = KafkaMessageType.valueOf(keyDeserializerName).toString();
ZiDeserializer<SolrInputDocument> deserializer = ZiDeserializerFactory.getInstance(validatedKeyDeserializerName);
result.setValueDeserializer(deserializer);
return result;
}
#Profile("!test")
#Bean
#Autowired
public ConcurrentKafkaListenerContainerFactory<String, SolrInputDocument> batchFactory(ConsumerFactory<String, SolrInputDocument> consumerFactory) {
ConcurrentKafkaListenerContainerFactory<String, SolrInputDocument> factory = new ConcurrentKafkaListenerContainerFactory<>();
factory.setConsumerFactory(consumerFactory);
factory.setBatchListener(true);
factory.setConcurrency(2);
ExponentialBackOffWithMaxRetries backoff = new ExponentialBackOffWithMaxRetries(10);
backoff.setMultiplier(3); // Default is 1.5 but this seems more reasonable
factory.setCommonErrorHandler(new DefaultErrorHandler(null, backoff));
// Needed for manual commits
factory.getContainerProperties().setAckMode(ContainerProperties.AckMode.MANUAL_IMMEDIATE);
return factory;
}
Note that the interface ZiDeserializer<SolrInputDocument> deserializeris my interface and ZiDeserializerFactory.getInstance(validatedKeyDeserializerName); returns my custom implementation of ZiDeserializer. And ZiDeserializer extends org.apache.kafka.common.serialization.Deserializer. This works for me
I have an application using spring boot & spring kafka that gets the delivery attempt header in a record interceptor so that I can include it in log messages. It has been working well until I upgraded to spring boot 2.6.3 and spring kafka 2.8.2 (from 2.5.5/2.7.7)
Now when I try to read the delivery attempt header it is not available. If I try and do the exact same thing within a message listener then it works fine so the header is clearly there.
This is what a simplified record interceptor and the listener container factory look like:
#Bean
public RecordInterceptor<Object, Object> recordInterceptor() {
return record -> {
int delivery = ByteBuffer.wrap(record.headers().lastHeader(KafkaHeaders.DELIVERY_ATTEMPT).value()).getInt();
log.info("delivery " + delivery);
return record;
};
}
#Bean
public ConcurrentKafkaListenerContainerFactory<?, ?> kafkaListenerContainerFactory(
ConcurrentKafkaListenerContainerFactoryConfigurer configurer,
ConsumerFactory<Object, Object> kafkaConsumerFactory) {
ConcurrentKafkaListenerContainerFactory<Object, Object> factory = new ConcurrentKafkaListenerContainerFactory<>();
configurer.configure(factory, kafkaConsumerFactory);
factory.getContainerProperties().setDeliveryAttemptHeader(true);
factory.setRecordInterceptor(recordInterceptor());
return factory;
}
I can't see anything in the spring docs suggesting the behaviour should have changed. Any ideas?
This is a bug; I opened an issue. https://github.com/spring-projects/spring-kafka/issues/2082
I've a JMS listener similar to the one done here:
Synchronous Message send and receive using JMS template and Spring Boot
It receives a XML payload. But different from that one, I want to also return a reply. If I return a String, it works ok, but as I'm doing with the message payload, I want to return a JAXB object.
However, if I return a Message, Spring tries to convert the object using the SimpleMessageConverter.
How can I configure the MarshallingMessageConverter to be used when converting the reply payload?
I had to configure the DefaultJmsListenerContainerFactory's message converter.
But what is weird, I already have a MarshallingMessageConverter there for DefaultMessageHandlerMethodFactory and JmsMessagingTemplate, and it's from another Java package (org.springframework.messaging.converter.MarshallingMessageConverter).
#Bean
public JmsListenerContainerFactory<?> jmsListenerFactory(ConnectionFactory connectionFactory,
DefaultJmsListenerContainerFactoryConfigurer configurer)
{
DefaultJmsListenerContainerFactory factory = new DefaultJmsListenerContainerFactory();
factory.setErrorHandler(errorHandler());
factory.setMessageConverter(jmsMarshallingMessageConverter()); // !!!!
configurer.configure(factory, connectionFactory);
return factory;
}
#Bean
public org.springframework.jms.support.converter.MarshallingMessageConverter jmsMarshallingMessageConverter()
{
Jaxb2Marshaller marshaller = marshaller();
org.springframework.jms.support.converter.MarshallingMessageConverter converter =
new org.springframework.jms.support.converter.MarshallingMessageConverter();
converter.setMarshaller(marshaller);
converter.setUnmarshaller(marshaller);
converter.setTargetType(MessageType.TEXT);
return converter;
}
I have successfully integrated kafka outbound channle adapter with fixed topic name. Now, i want to make the topic name configurable and hence, want to expose it via application properties.
application.properties contain one of the following entry:
kafkaTopic:testNewTopic
My configuration class looks like below:
#Configuration
#Component
public class KafkaConfig {
#Value("${kafkaTopic}")
private String kafkaTopicName;
#Bean
public String getTopic(){
return kafkaTopicName;
}
#Bean
public KafkaTemplate<String, String> kafkaTemplate() {
return new KafkaTemplate<>(producerFactory());
}
#Bean
public ProducerFactory<String, String> producerFactory() {
Map<String, Object> props = new HashMap<>();
props.put(ProducerConfig.BOOTSTRAP_SERVERS_CONFIG, "localhost:9092");//this.brokerAddress);
props.put(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG, "org.apache.kafka.common.serialization.StringSerializer");
props.put(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG, "org.apache.kafka.common.serialization.StringSerializer");
// set more properties
return new DefaultKafkaProducerFactory<>(props);
}
}
and in my si-config.xml, i have used the following (ex: topic="getTopic") :
<int-kafka:outbound-channel-adapter
id="kafkaOutboundChannelAdapter" kafka-template="kafkaTemplate"
auto-startup="true" sync="true" channel="inputToKafka" topic="getTopic">
</int-kafka:outbound-channel-adapter>
However, the configuration is unable to pick up the topic name when exposed via bean. But it works fine when i hard code the value of the topic name.
Can someone please suggest what i am doing wrong here?
Does topic within kafka outbound channel accept the value referred as bean?
How do i externalize it as every application using my utility will supply different kafka topic names
The topic attribute is for string value.
However it supports property placeholder resolution:
topic="${kafkaTopic}"
and also SpEL evaluation for aforementioned bean:
topic="#{getTopic}"
Just because this is allowed by the XML parser configuration.
However you may pay attention that KafkaTemplate, which you inject into the <int-kafka:outbound-channel-adapter> has defaultTopic property. Therefore you won't need to worry about that XML.
And one more option available for you is Spring Integration Annotations configuration. Where you can define a #ServiceActivator for the KafkaProducerMessageHandler #Bean:
#ServiceActivator(inputChannel = "inputToKafka")
#Bean
KafkaProducerMessageHandler kafkaOutboundChannelAdapter() {
kafkaOutboundChannelAdapter adapter = new kafkaOutboundChannelAdapter( kafkaTemplate());
adapter.setSync(true);
adapter.setTopicExpression(new LiteralExpression(this.kafkaTopicName));
return adapter;
}
I am trying to dynamically resolve;
#JmsListener(destinations = "${some.key}")
Now I know we can resolve this using properties using such thing as PropertySourcesPlaceholderConfigurer.
I have found the following spring ticket, https://jira.spring.io/browse/SPR-12289. Which gives me some indication that this can be done without the use of properties files using a DestinationResolver. I am hoping that this is a resolver that is not property based but I am not 100% sure. Could some give me an example of this without using xml based spring.
See the documentation which shows how to configure a custom destination resolver in the listener container factory that generates the listener for the annotation:
#Bean
public DefaultJmsListenerContainerFactory jmsListenerContainerFactory() {
DefaultJmsListenerContainerFactory factory =
new DefaultJmsListenerContainerFactory();
factory.setConnectionFactory(connectionFactory());
factory.setDestinationResolver(destinationResolver());
factory.setConcurrency("3-10");
return factory;
}