I am trying to write kafka consumer using spring-kafka version 2.3.0.M2 library.
To handle run time errors I am using SeekToCurrentErrorHandler.class with DeadLetterPublishingRecoverer as my recoverer. This works fine only when my consumer code throws exception, but fails when unable to deserialize the message.
I tried implementing ErrorHandler myself and I was successful but with this approach I myself end up writing DLT code to handle error messages which I do not want to do.
Below are my kafka properties
spring:
kafka:
consumer:
bootstrap-servers: localhost:9092
group-id: group_id
auto-offset-reset: latest
key-deserializer: org.springframework.kafka.support.serializer.ErrorHandlingDeserializer2
value-deserializer: org.springframework.kafka.support.serializer.ErrorHandlingDeserializer2
properties:
spring.json.trusted.packages: com.mypackage
spring.deserializer.key.delegate.class: org.apache.kafka.common.serialization.StringDeserializer
spring.deserializer.value.delegate.class: org.apache.kafka.common.serialization.StringDeserializer
public ConcurrentKafkaListenerContainerFactory kafkaListenerContainerFactory(
ConcurrentKafkaListenerContainerFactoryConfigurer configurer,
ConsumerFactory<Object, Object> kafkaConsumerFactory,
KafkaTemplate<Object, Object> template) {
ConcurrentKafkaListenerContainerFactory<Object, Object> factory = new ConcurrentKafkaListenerContainerFactory<>();
configurer.configure(factory, kafkaConsumerFactory);
factory.setErrorHandler(new SeekToCurrentErrorHandler(new DeadLetterPublishingRecoverer(template), maxFailures));}
It works fine for me (note that Boot will auto-configure the error handler)...
#SpringBootApplication
public class So56728833Application {
public static void main(String[] args) {
SpringApplication.run(So56728833Application.class, args);
}
#Bean
public SeekToCurrentErrorHandler errorHandler(KafkaTemplate<String, String> template) {
SeekToCurrentErrorHandler eh = new SeekToCurrentErrorHandler(new DeadLetterPublishingRecoverer(template), 3);
eh.setClassifier( // retry for all except deserialization exceptions
new BinaryExceptionClassifier(Collections.singletonList(DeserializationException.class), false));
return eh;
}
#KafkaListener(id = "so56728833"
+ "", topics = "so56728833")
public void listen(Foo in) {
System.out.println(in);
if (in.getBar().equals("baz")) {
throw new IllegalStateException("Test retries");
}
}
#KafkaListener(id = "so56728833dlt", topics = "so56728833.DLT")
public void listenDLT(Object in) {
System.out.println("Received from DLT: " + (in instanceof byte[] ? new String((byte[]) in) : in));
}
#Bean
public NewTopic topic() {
return TopicBuilder.name("so56728833").partitions(1).replicas(1).build();
}
#Bean
public NewTopic dlt() {
return TopicBuilder.name("so56728833.DLT").partitions(1).replicas(1).build();
}
public static class Foo {
private String bar;
public Foo() {
super();
}
public Foo(String bar) {
this.bar = bar;
}
public String getBar() {
return this.bar;
}
public void setBar(String bar) {
this.bar = bar;
}
#Override
public String toString() {
return "Foo [bar=" + this.bar + "]";
}
}
}
spring:
kafka:
consumer:
auto-offset-reset: earliest
enable-auto-commit: false
key-deserializer: org.springframework.kafka.support.serializer.ErrorHandlingDeserializer2
value-deserializer: org.springframework.kafka.support.serializer.ErrorHandlingDeserializer2
properties:
spring.json.trusted.packages: com.example
spring.deserializer.key.delegate.class: org.springframework.kafka.support.serializer.JsonDeserializer
spring.deserializer.value.delegate.class: org.springframework.kafka.support.serializer.JsonDeserializer
spring.json.value.default.type: com.example.So56728833Application$Foo
producer:
key-serializer: org.springframework.kafka.support.serializer.JsonSerializer
value-serializer: org.springframework.kafka.support.serializer.JsonSerializer
logging:
level:
org.springframework.kafka: trace
I have 3 records in the topic:
"badJSON"
"{\"bar\":\"baz\"}"
"{\"bar\":\"qux\"}"
I see the first one going directly to the DLT, and the second one goes there after 3 attempts.
Related
Just wanted to make sure that whether messages are processed in correct way or not. When the message gets received at listener, it will be always processed by a new thread( defined the processor bean as prototype). is this implementation correct ? (i have Considered the listener is not thread safe, so for this reason the prototype scope of bean to process the message has been used)
(Input : TestTopic- 5 partitions - 1 consumer) or (Input : TestTopic- 5 partitions - 5 consumers)
public class EventListener {
#Autowired
private EventProcessor eventProcessor;
#KafkaListener(topics = "TestTopic", containerFactory = "kafkaListenerContainerFactory",
autoStartup = "true")
public void onMessage(
#Payload List<ConsumerRecord<String, String>> consumerRecords, Acknowledgment acknowledgment) {
eventProcessor.processAndAcknowledgeBatchMessages(consumerRecords, acknowledgment);
}
}
//event processor
#Slf4j
#Component
#Scope(ConfigurableBeanFactory.SCOPE_PROTOTYPE)
#NoArgsConstructor
#SuppressWarnings("unused")
public class EventProcessorImpl implements EventProcessor {
#Autowired
private KafkaProducerTemplate kafkaProducerTemplate;
#Autowired
private ObjectMapper localObjectMapper;
#Autowired
private Dao dao;
public void processAndAcknowledgeBatchMessages(
List<ConsumerRecord<String, String>> consumerRecords, Acknowledgment acknowledgment) {
long start = System.currentTimeMillis();
consumerRecords.forEach( consumerRecord -> {
Event event = localObjectMapper.readValue(consumerRecord.value(), Event.class);
dao.save(process(event));
});
acknowledgment.acknowledge();
}
}
No it is not correct; you should not execute on another thread; it will cause problems with committing offsets and error handling.
Also, making the EventProcessorImpl a prototype bean won't help. That just means a new instance is used each time the bean is referenced.
Since it is #Autowired it is only referenced once, during initialization. To get a new instance for each request, you would need to call getBean() on the application context each time.
It is better to make your code thread-safe.
EDIT
There are (at least) a couple of ways to deal with a not thread-safe service defined in prototype scope.
Use a ThreadLocal:
#SpringBootApplication
public class So68447863Application {
public static void main(String[] args) {
SpringApplication.run(So68447863Application.class, args);
}
private static final ThreadLocal<NotThreadSafeService> SERVICES = new ThreadLocal<>();
#Autowired
ApplicationContext context;
#KafkaListener(id = "so68447863", topics = "so68447863", concurrency = "5")
void listen(String in) {
NotThreadSafeService service = SERVICES.get();
if (service == null) {
service = this.context.getBean(NotThreadSafeService.class);
SERVICES.set(service);
}
service.process(in);
}
#EventListener
void removeService(ConsumerStoppedEvent event) {
System.out.println("Consumer stopped; removing TL");
SERVICES.remove();
}
#Bean
NewTopic topic() {
return TopicBuilder.name("so68447863").partitions(10).replicas(1).build();
}
#Bean
#Scope(ConfigurableBeanFactory.SCOPE_PROTOTYPE)
NotThreadSafeService service() {
return new NotThreadSafeService();
}
}
class NotThreadSafeService {
void process(String msg) {
System.out.println(msg + " processed by " + this);
}
}
Use a pool of instances.
#SpringBootApplication
public class So68447863Application {
public static void main(String[] args) {
SpringApplication.run(So68447863Application.class, args);
}
private static final BlockingQueue<NotThreadSafeService> SERVICES = new LinkedBlockingQueue<>();
#Autowired
ApplicationContext context;
#KafkaListener(id = "so68447863", topics = "so68447863", concurrency = "5")
void listen(String in) {
NotThreadSafeService service = SERVICES.poll();
if (service == null) {
service = this.context.getBean(NotThreadSafeService.class);
}
try {
service.process(in);
}
finally {
SERVICES.add(service);
}
}
#Bean
NewTopic topic() {
return TopicBuilder.name("so68447863").partitions(10).replicas(1).build();
}
#Bean
#Scope(ConfigurableBeanFactory.SCOPE_PROTOTYPE)
NotThreadSafeService service() {
return new NotThreadSafeService();
}
}
class NotThreadSafeService {
void process(String msg) {
System.out.println(msg + " processed by " + this);
}
}
I am working on the spring kafka batch listener filter strategy. I am facing an issue that, the filtered events are coming again and again. could any one help me on this issue ? spring boot with kafka version(2.3.8)
Here is my configuration:
ConcurrentKafkaListenerContainerFactory<Object, Object> factory = new
ConcurrentKafkaListenerContainerFactory<>();
configurer.configure(factory, kafkaConsumerFactory);
factory.setBatchListener(true);
factory.setAckDiscarded(true);
factory.getContainerProperties().setIdleBetweenPolls(30000);
factory.setRecordFilterStrategy(
(consumerRecord) -> {
MyObject myObject = new ObjectMapper().readValue(consumerRecord.value(), MyObj.class);
if (myObject.frequency > 10) {
return false;
} else {
return true;
}});
factory.setBatchErrorHandler(new SeekToCurrentBatchErrorHandler());
When using batch mode with MANUAL acks, if you filter all the records (discard them all), the listener will get an empty list so you can still acknowledge the batch to commit the offsets.
I just tested it and it works as expected.
#SpringBootApplication
public class So67259790Application {
public static void main(String[] args) {
SpringApplication.run(So67259790Application.class, args);
}
#KafkaListener(id = "so67259790", topics = "so67259790")
public void listen(List<String> in, Acknowledgment ack) {
System.out.println(in);
ack.acknowledge();
}
#Bean
public NewTopic topic() {
return TopicBuilder.name("so67259790").partitions(1).replicas(1).build();
}
#Bean
public ApplicationRunner runner(KafkaTemplate<String, String> template) {
return args -> {
template.send("so67259790", "foo");
template.send("so67259790", "bar");
};
}
#Bean
public RecordFilterStrategy<Object, Object> rfs() {
return rec -> true;
}
}
I'm using spring boot 2.1.7.RELEASE and spring-kafka 2.2.8.RELEASE.And I'm using #KafkaListener annotation to create a consumer and I'm using all default settings for the consumer.And I'm using below configuration as specified in the Spring-Kafka documentation.
// other props
props.put(ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG, ErrorHandlingDeserializer2.class);
props.put(ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG, ErrorHandlingDeserializer2.class);
props.put(ErrorHandlingDeserializer.KEY_DESERIALIZER_CLASS, StringDeserializer.class);
props.put(ErrorHandlingDeserializer.VALUE_DESERIALIZER_CLASS, AvroDeserializer.class.getName());
return new DefaultKafkaConsumerFactory<>(props);
Now, I've implemented my custom SeekToCurrentErrorHandler by extending SeekToCurrentErrorHandler as per the below thread but the record value is coming as null and the record key is not in a readable format. Please suggest me how can i get the record key and value?
How to capture the exception and message key when using ErrorHandlingDeserializer2 to handle exceptions during deserialization
Here is my custom SeekToCurrentErrorHandler code
#Component
public class MySeekToCurrentErrorHandler extends SeekToCurrentErrorHandler {
private final MyDeadLetterRecoverer deadLetterRecoverer;
#Autowired
public MySeekToCurrentErrorHandler(MyDeadLetterRecoverer deadLetterRecoverer) {
super(-1);
this.deadLetterRecoverer = deadLetterRecoverer;
}
#Override
public void handle(Exception thrownException, List<ConsumerRecord<?, ?>> data, Consumer<?, ?> consumer, MessageListenerContainer container) {
if (thrownException instanceof DeserializationException) {
//Improve to support multiple records
DeserializationException deserializationException = (DeserializationException) thrownException;
deadLetterRecoverer.accept(data.get(0), deserializationException);
ConsumerRecord<?, ?>. consumerRecord = data.get(0);
sout(consumerRecord.key());
sout(consumerRecord.value());
} else {
//Calling super method to let the 'SeekToCurrentErrorHandler' do what it is actually designed for
super.handle(thrownException, data, consumer, container);
}
}
}
If the key fails deserialization, the original byte[] can be obtained by calling getData() on the exception.
Similarly, if the value fails deserialization, use getData() to get the original data.
The DeadLetterPublishingRecoverer does this (since 2.3).
You can tell which of the key or value failed by calling isKey() on the exception.
EDIT
I was wrong, the key and value are available if the value or key failed.
This is written with Boot 2.3.4:
#SpringBootApplication
public class Application {
public static void main(String[] args) {
SpringApplication.run(Application.class, args);
}
#Bean
SeekToCurrentErrorHandler errorHandler(ProducerFactory<String, String> pf) {
Map<String, Object> configs = new HashMap<>(pf.getConfigurationProperties());
configs.put(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG, ByteArraySerializer.class);
configs.put(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG, ByteArraySerializer.class);
ProducerFactory<byte[], byte[]> bytesPF = new DefaultKafkaProducerFactory<>(configs);
KafkaOperations<byte[], byte[]> template = new KafkaTemplate<>(bytesPF);
return new SeekToCurrentErrorHandler(new DeadLetterPublishingRecoverer(template),
new FixedBackOff(1000, 5));
}
#KafkaListener(id = "so64597061", topics = "so64597061",
properties = {
ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG
+ ":org.springframework.kafka.support.serializer.ErrorHandlingDeserializer",
ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG
+ ":org.springframework.kafka.support.serializer.ErrorHandlingDeserializer",
ErrorHandlingDeserializer.KEY_DESERIALIZER_CLASS
+ ":com.example.demo.Application$FailSometimesDeserializer",
ErrorHandlingDeserializer.VALUE_DESERIALIZER_CLASS
+ ":com.example.demo.Application$FailSometimesDeserializer"
})
public void listen(String val, #Header(name = KafkaHeaders.RECEIVED_MESSAGE_KEY) String key) {
System.out.println(key + ":" + val);
}
#KafkaListener(id = "so64597061.dlt", topics = "so64597061.DLT",
properties = {
ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG
+ ":org.apache.kafka.common.serialization.ByteArrayDeserializer",
ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG
+ ":org.apache.kafka.common.serialization.ByteArrayDeserializer"
})
public void dltListen(byte[] val, #Header(name = KafkaHeaders.RECEIVED_MESSAGE_KEY, required = false) byte[] key) {
String keyStr = key != null ? new String(key) : null;
String valStr = val != null ? new String(val) : null;
System.out.println("DLT:" + keyStr + ":" + valStr);
}
#Bean
public ApplicationRunner runner(KafkaTemplate<String, String> template) {
return args -> {
template.send("so64597061", "foo", "bar");
template.send("so64597061", "fail", "keyFailed");
template.send("so64597061", "valueFailed", "fail");
};
}
#Bean
public NewTopic topic() {
return TopicBuilder.name("so64597061").partitions(1).replicas(1).build();
}
#Bean
public NewTopic dlt() {
return TopicBuilder.name("so64597061.DLT").partitions(1).replicas(1).build();
}
public static class FailSometimesDeserializer implements Deserializer<byte[]> {
#Override
public void configure(Map<String, ?> configs, boolean isKey) {
}
#Override
public byte[] deserialize(String topic, byte[] data) {
return data;
}
#Override
public void close() {
}
#Override
public byte[] deserialize(String topic, Headers headers, byte[] data) {
String string = new String(data);
if ("fail".equals(string)) {
throw new RuntimeException("fail");
}
return data;
}
}
}
spring.kafka.consumer.auto-offset-reset=earliest
foo:bar
DLT:fail:keyFailed
DLT:valueFailed:fail
I'd like to use a Kafka state store of type KeyValueStore in a sample application using the Kafka Binder of Spring Cloud Stream.
According to the documentation, it should be pretty simple.
This is my main class:
#SpringBootApplication
public class KafkaStreamTestApplication {
public static void main(String[] args) {
SpringApplication.run(KafkaStreamTestApplication.class, args);
}
#Bean
public BiFunction<KStream<String, String>, KeyValueStore<String,String>, KStream<String, String>> process(){
return (input,store) -> input.mapValues(v -> v.toUpperCase());
}
#Bean
public StoreBuilder myStore() {
return Stores.keyValueStoreBuilder(
Stores.persistentKeyValueStore("my-store"), Serdes.String(),
Serdes.String());
}
}
I suppose that the KeyValueStore should be passed as the second parameter of the "process" method, but the application fails to start with the message below:
Caused by: java.lang.IllegalStateException: No factory found for binding target type: org.apache.kafka.streams.state.KeyValueStore among registered factories: channelFactory,messageSourceFactory,kStreamBoundElementFactory,kTableBoundElementFactory,globalKTableBoundElementFactory
at org.springframework.cloud.stream.binding.AbstractBindableProxyFactory.getBindingTargetFactory(AbstractBindableProxyFactory.java:82) ~[spring-cloud-stream-3.0.3.RELEASE.jar:3.0.3.RELEASE]
at org.springframework.cloud.stream.binder.kafka.streams.function.KafkaStreamsBindableProxyFactory.bindInput(KafkaStreamsBindableProxyFactory.java:191) ~[spring-cloud-stream-binder-kafka-streams-3.0.3.RELEASE.jar:3.0.3.RELEASE]
at org.springframework.cloud.stream.binder.kafka.streams.function.KafkaStreamsBindableProxyFactory.afterPropertiesSet(KafkaStreamsBindableProxyFactory.java:103) ~[spring-cloud-stream-binder-kafka-streams-3.0.3.RELEASE.jar:3.0.3.RELEASE]
at org.springframework.beans.factory.support.AbstractAutowireCapableBeanFactory.invokeInitMethods(AbstractAutowireCapableBeanFactory.java:1855) ~[spring-beans-5.2.5.RELEASE.jar:5.2.5.RELEASE]
at org.springframework.beans.factory.support.AbstractAutowireCapableBeanFactory.initializeBean(AbstractAutowireCapableBeanFactory.java:1792) ~[spring-beans-5.2.5.RELEASE.jar:5.2.5.RELEASE]
I found the solution about how to use a store reading an unit test in Spring Cloud Stream.
The code below is how I applied that solution to my code.
The transformer uses the Store provided by Spring bean method "myStore"
#SpringBootApplication
public class KafkaStreamTestApplication {
public static final String MY_STORE_NAME = "my-store";
public static void main(String[] args) {
SpringApplication.run(KafkaStreamTestApplication.class, args);
}
#Bean
public Function<KStream<String, String>, KStream<String, String>> process2(){
return (input) -> input.
transformValues(() -> new MyValueTransformer(), MY_STORE_NAME);
}
#Bean
public StoreBuilder<?> myStore() {
return Stores.keyValueStoreBuilder(
Stores.persistentKeyValueStore(MY_STORE_NAME), Serdes.String(),
Serdes.String());
}
}
public class MyValueTransformer implements ValueTransformer<String, String> {
private KeyValueStore<String,String> store;
private ProcessorContext context;
#Override
public void init(ProcessorContext context) {
this.context = context;
store = (KeyValueStore<String, String>) this.context.getStateStore(KafkaStreamTestApplication.MY_STORE_NAME);
}
#Override
public String transform(String value) {
String tValue = store.get(value);
if(tValue==null) {
store.put(value, value.toUpperCase());
}
return tValue;
}
#Override
public void close() {
if(store!=null) {
store.close();
}
}
}
Is it possible to use ReplyingKafkaTemplate with Spring Cloud Stream? Is there any code example of a configuration in order to use it?
This is all in one application but illustrates how it works...
#SpringBootApplication
#EnableBinding(Processor.class)
public class So57380643Application {
public static void main(String[] args) {
SpringApplication.run(So57380643Application.class, args).close();
}
#Bean
public ReplyingKafkaTemplate<byte[], byte[], byte[]> replyer(ProducerFactory<byte[], byte[]> pf,
ConcurrentMessageListenerContainer<byte[], byte[]> replyContainer) {
return new ReplyingKafkaTemplate<>(pf, replyContainer);
}
#Bean
public ConcurrentMessageListenerContainer<byte[], byte[]> replyContainer(
ConcurrentKafkaListenerContainerFactory<byte[], byte[]> factory) {
ConcurrentMessageListenerContainer<byte[], byte[]> container = factory.createContainer("replyTopic");
container.getContainerProperties().setGroupId("replies.group");
return container;
}
#StreamListener(Processor.INPUT)
#SendTo(Processor.OUTPUT)
public String listen(String in) {
return in.toUpperCase();
}
#Bean
public ApplicationRunner runner(ReplyingKafkaTemplate<byte[], byte[], byte[]> replyer) {
return args -> {
ProducerRecord<byte[], byte[]> record = new ProducerRecord<>("requestTopic", "foo".getBytes());
RequestReplyFuture<byte[], byte[], byte[]> future = replyer.sendAndReceive(record);
RecordMetadata meta = future.getSendFuture().get(10, TimeUnit.SECONDS).getRecordMetadata();
System.out.println(meta);
ConsumerRecord<byte[], byte[]> consumerRecord = future.get(10, TimeUnit.SECONDS);
System.out.println(new String(consumerRecord.value()));
};
}
}
and
spring:
kafka:
consumer:
enable-auto-commit: false
auto-offset-reset: earliest
cloud:
stream:
bindings:
input:
destination: requestTopic
group: so57380643
output:
destination: replyTopic
result:
requestTopic-0#3
FOO