40. Apache Kafka Streams Binder

40.1 Usage

For using the Kafka Streams binder, you just need to add it to your Spring Cloud Stream application, using the following Maven coordinates:

<dependency>
  <groupId>org.springframework.cloud</groupId>
  <artifactId>spring-cloud-stream-binder-kafka-streams</artifactId>
</dependency>

40.2 Kafka Streams Binder Overview

Spring Cloud Stream’s Apache Kafka support also includes a binder implementation designed explicitly for Apache Kafka Streams binding. With this native integration, a Spring Cloud Stream "processor" application can directly use the Apache Kafka Streams APIs in the core business logic.

Kafka Streams binder implementation builds on the foundation provided by the Kafka Streams in Spring Kafka project.

Kafka Streams binder provides binding capabilities for the three major types in Kafka Streams - KStream, KTable and GlobalKTable.

As part of this native integration, the high-level Streams DSL provided by the Kafka Streams API is available for use in the business logic.

An early version of the Processor API support is available as well.

As noted early-on, Kafka Streams support in Spring Cloud Stream is strictly only available for use in the Processor model. A model in which the messages read from an inbound topic, business processing can be applied, and the transformed messages can be written to an outbound topic. It can also be used in Processor applications with a no-outbound destination.

40.2.1 Streams DSL

This application consumes data from a Kafka topic (e.g., words), computes word count for each unique word in a 5 seconds time window, and the computed results are sent to a downstream topic (e.g., counts) for further processing.

@SpringBootApplication
@EnableBinding(KStreamProcessor.class)
public class WordCountProcessorApplication {

	@StreamListener("input")
	@SendTo("output")
	public KStream<?, WordCount> process(KStream<?, String> input) {
		return input
                .flatMapValues(value -> Arrays.asList(value.toLowerCase().split("\\W+")))
                .groupBy((key, value) -> value)
                .windowedBy(TimeWindows.of(5000))
                .count(Materialized.as("WordCounts-multi"))
                .toStream()
                .map((key, value) -> new KeyValue<>(null, new WordCount(key.key(), value, new Date(key.window().start()), new Date(key.window().end()))));
    }

	public static void main(String[] args) {
		SpringApplication.run(WordCountProcessorApplication.class, args);
	}

Once built as a uber-jar (e.g., wordcount-processor.jar), you can run the above example like the following.

java -jar wordcount-processor.jar  --spring.cloud.stream.bindings.input.destination=words --spring.cloud.stream.bindings.output.destination=counts

This application will consume messages from the Kafka topic words and the computed results are published to an output topic counts.

Spring Cloud Stream will ensure that the messages from both the incoming and outgoing topics are automatically bound as KStream objects. As a developer, you can exclusively focus on the business aspects of the code, i.e. writing the logic required in the processor. Setting up the Streams DSL specific configuration required by the Kafka Streams infrastructure is automatically handled by the framework.

40.3 Configuration Options

This section contains the configuration options used by the Kafka Streams binder.

For common configuration options and properties pertaining to binder, refer to the core documentation.

40.3.1 Kafka Streams Properties

The following properties are available at the binder level and must be prefixed with spring.cloud.stream.kafka.streams.binder. literal.

configuration
Map with a key/value pair containing properties pertaining to Apache Kafka Streams API. This property must be prefixed with spring.cloud.stream.kafka.streams.binder.. Following are some examples of using this property.
spring.cloud.stream.kafka.streams.binder.configuration.default.key.serde=org.apache.kafka.common.serialization.Serdes$StringSerde
spring.cloud.stream.kafka.streams.binder.configuration.default.value.serde=org.apache.kafka.common.serialization.Serdes$StringSerde
spring.cloud.stream.kafka.streams.binder.configuration.commit.interval.ms=1000

For more information about all the properties that may go into streams configuration, see StreamsConfig JavaDocs in Apache Kafka Streams docs.

brokers

Broker URL

Default: localhost

zkNodes

Zookeeper URL

Default: localhost

serdeError

Deserialization error handler type. Possible values are - logAndContinue, logAndFail or sendToDlq

Default: logAndFail

applicationId

Convenient way to set the application.id for the Kafka Streams application globally at the binder level. If the application contains multiple StreamListener methods, then application.id should be set at the binding level per input binding.

Default: none

The following properties are only available for Kafka Streams producers and must be prefixed with spring.cloud.stream.kafka.streams.bindings.<binding name>.producer. literal. For convenience, if there multiple output bindings and they all require a common value, that can be configured by using the prefix spring.cloud.stream.kafka.streams.default.producer..

keySerde

key serde to use

Default: none.

valueSerde

value serde to use

Default: none.

useNativeEncoding

flag to enable native encoding

Default: false.

The following properties are only available for Kafka Streams consumers and must be prefixed with spring.cloud.stream.kafka.streams.bindings.<binding name>.consumer.`literal. For convenience, if there multiple input bindings and they all require a common value, that can be configured by using the prefix `spring.cloud.stream.kafka.streams.default.consumer..

applicationId

Setting application.id per input binding.

Default: none

keySerde

key serde to use

Default: none.

valueSerde

value serde to use

Default: none.

materializedAs

state store to materialize when using incoming KTable types

Default: none.

useNativeDecoding

flag to enable native decoding

Default: false.

dlqName

DLQ topic name.

Default: none.

40.3.2 TimeWindow properties:

Windowing is an important concept in stream processing applications. Following properties are available to configure time-window computations.

spring.cloud.stream.kafka.streams.timeWindow.length

When this property is given, you can autowire a TimeWindows bean into the application. The value is expressed in milliseconds.

Default: none.

spring.cloud.stream.kafka.streams.timeWindow.advanceBy

Value is given in milliseconds.

Default: none.

40.4 Multiple Input Bindings

For use cases that requires multiple incoming KStream objects or a combination of KStream and KTable objects, the Kafka Streams binder provides multiple bindings support.

Let’s see it in action.

40.4.1 Multiple Input Bindings as a Sink

@EnableBinding(KStreamKTableBinding.class)
.....
.....
@StreamListener
public void process(@Input("inputStream") KStream<String, PlayEvent> playEvents,
                    @Input("inputTable") KTable<Long, Song> songTable) {
                    ....
                    ....
}

interface KStreamKTableBinding {

    @Input("inputStream")
    KStream<?, ?> inputStream();

    @Input("inputTable")
    KTable<?, ?> inputTable();
}

In the above example, the application is written as a sink, i.e. there are no output bindings and the application has to decide concerning downstream processing. When you write applications in this style, you might want to send the information downstream or store them in a state store (See below for Queryable State Stores).

In the case of incoming KTable, if you want to materialize the computations to a state store, you have to express it through the following property.

spring.cloud.stream.kafka.streams.bindings.inputTable.consumer.materializedAs: all-songs

The above example shows the use of KTable as an input binding. The binder also supports input bindings for GlobalKTable. GlobalKTable binding is useful when you have to ensure that all instances of your application has access to the data updates from the topic. KTable and GlobalKTable bindings are only available on the input. Binder supports both input and output bindings for KStream.

40.4.2 Multiple Input Bindings as a Processor

@EnableBinding(KStreamKTableBinding.class)
....
....

@StreamListener
@SendTo("output")
public KStream<String, Long> process(@Input("input") KStream<String, Long> userClicksStream,
                                     @Input("inputTable") KTable<String, String> userRegionsTable) {
....
....
}

interface KStreamKTableBinding extends KafkaStreamsProcessor {

    @Input("inputX")
    KTable<?, ?> inputTable();
}

40.5 Multiple Output Bindings (aka Branching)

Kafka Streams allow outbound data to be split into multiple topics based on some predicates. The Kafka Streams binder provides support for this feature without compromising the programming model exposed through StreamListener in the end user application.

You can write the application in the usual way as demonstrated above in the word count example. However, when using the branching feature, you are required to do a few things. First, you need to make sure that your return type is KStream[] instead of a regular KStream. Second, you need to use the SendTo annotation containing the output bindings in the order (see example below). For each of these output bindings, you need to configure destination, content-type etc., complying with the standard Spring Cloud Stream expectations.

Here is an example:

@EnableBinding(KStreamProcessorWithBranches.class)
@EnableAutoConfiguration
public static class WordCountProcessorApplication {

    @Autowired
    private TimeWindows timeWindows;

    @StreamListener("input")
    @SendTo({"output1","output2","output3})
    public KStream<?, WordCount>[] process(KStream<Object, String> input) {

			Predicate<Object, WordCount> isEnglish = (k, v) -> v.word.equals("english");
			Predicate<Object, WordCount> isFrench =  (k, v) -> v.word.equals("french");
			Predicate<Object, WordCount> isSpanish = (k, v) -> v.word.equals("spanish");

			return input
					.flatMapValues(value -> Arrays.asList(value.toLowerCase().split("\\W+")))
					.groupBy((key, value) -> value)
					.windowedBy(timeWindows)
					.count(Materialized.as("WordCounts-1"))
					.toStream()
					.map((key, value) -> new KeyValue<>(null, new WordCount(key.key(), value, new Date(key.window().start()), new Date(key.window().end()))))
					.branch(isEnglish, isFrench, isSpanish);
    }

    interface KStreamProcessorWithBranches {

    		@Input("input")
    		KStream<?, ?> input();

    		@Output("output1")
    		KStream<?, ?> output1();

    		@Output("output2")
    		KStream<?, ?> output2();

    		@Output("output3")
    		KStream<?, ?> output3();
    	}
}

Properties:

spring.cloud.stream.bindings.output1.contentType: application/json
spring.cloud.stream.bindings.output2.contentType: application/json
spring.cloud.stream.bindings.output3.contentType: application/json
spring.cloud.stream.kafka.streams.binder.configuration.commit.interval.ms: 1000
spring.cloud.stream.kafka.streams.binder.configuration:
  default.key.serde: org.apache.kafka.common.serialization.Serdes$StringSerde
  default.value.serde: org.apache.kafka.common.serialization.Serdes$StringSerde
spring.cloud.stream.bindings.output1:
  destination: foo
  producer:
    headerMode: raw
spring.cloud.stream.bindings.output2:
  destination: bar
  producer:
    headerMode: raw
spring.cloud.stream.bindings.output3:
  destination: fox
  producer:
    headerMode: raw
spring.cloud.stream.bindings.input:
  destination: words
  consumer:
    headerMode: raw

40.6 Message Conversion

Similar to message-channel based binder applications, the Kafka Streams binder adapts to the out-of-the-box content-type conversions without any compromise.

It is typical for Kafka Streams operations to know the type of SerDe’s used to transform the key and value correctly. Therefore, it may be more natural to rely on the SerDe facilities provided by the Apache Kafka Streams library itself at the inbound and outbound conversions rather than using the content-type conversions offered by the framework. On the other hand, you might be already familiar with the content-type conversion patterns provided by the framework, and that, you’d like to continue using for inbound and outbound conversions.

Both the options are supported in the Kafka Streams binder implementation.

40.6.1 Outbound serialization

If native encoding is disabled (which is the default), then the framework will convert the message using the contentType set by the user (otherwise, the default application/json will be applied). It will ignore any SerDe set on the outbound in this case for outbound serialization.

Here is the property to set the contentType on the outbound.

spring.cloud.stream.bindings.output.contentType: application/json

Here is the property to enable native encoding.

spring.cloud.stream.bindings.output.nativeEncoding: true

If native encoding is enabled on the output binding (user has to enable it as above explicitly), then the framework will skip any form of automatic message conversion on the outbound. In that case, it will switch to the Serde set by the user. The valueSerde property set on the actual output binding will be used. Here is an example.

spring.cloud.stream.kafka.streams.bindings.output.producer.valueSerde: org.apache.kafka.common.serialization.Serdes$StringSerde

If this property is not set, then it will use the "default" SerDe: spring.cloud.stream.kafka.streams.binder.configuration.default.value.serde.

It is worth to mention that Kafka Streams binder does not serialize the keys on outbound - it simply relies on Kafka itself. Therefore, you either have to specify the keySerde property on the binding or it will default to the application-wide common keySerde.

Binding level key serde:

spring.cloud.stream.kafka.streams.bindings.output.producer.keySerde

Common Key serde:

spring.cloud.stream.kafka.streams.binder.configuration.default.key.serde

If branching is used, then you need to use multiple output bindings. For example,

interface KStreamProcessorWithBranches {

    		@Input("input")
    		KStream<?, ?> input();

    		@Output("output1")
    		KStream<?, ?> output1();

    		@Output("output2")
    		KStream<?, ?> output2();

    		@Output("output3")
    		KStream<?, ?> output3();
    	}

If nativeEncoding is set, then you can set different SerDe’s on individual output bindings as below.

spring.cloud.stream.kafka.streams.bindings.output1.producer.valueSerde=IntegerSerde
spring.cloud.stream.kafka.streams.bindings.output2.producer.valueSerde=StringSerde
spring.cloud.stream.kafka.streams.bindings.output3.producer.valueSerde=JsonSerde

Then if you have SendTo like this, @SendTo({"output1", "output2", "output3"}), the KStream[] from the branches are applied with proper SerDe objects as defined above. If you are not enabling nativeEncoding, you can then set different contentType values on the output bindings as below. In that case, the framework will use the appropriate message converter to convert the messages before sending to Kafka.

spring.cloud.stream.bindings.output1.contentType: application/json
spring.cloud.stream.bindings.output2.contentType: application/java-serialzied-object
spring.cloud.stream.bindings.output3.contentType: application/octet-stream

40.6.2 Inbound Deserialization

Similar rules apply to data deserialization on the inbound.

If native decoding is disabled (which is the default), then the framework will convert the message using the contentType set by the user (otherwise, the default application/json will be applied). It will ignore any SerDe set on the inbound in this case for inbound deserialization.

Here is the property to set the contentType on the inbound.

spring.cloud.stream.bindings.input.contentType: application/json

Here is the property to enable native decoding.

spring.cloud.stream.bindings.input.nativeDecoding: true

If native decoding is enabled on the input binding (user has to enable it as above explicitly), then the framework will skip doing any message conversion on the inbound. In that case, it will switch to the SerDe set by the user. The valueSerde property set on the actual output binding will be used. Here is an example.

spring.cloud.stream.kafka.streams.bindings.input.consumer.valueSerde: org.apache.kafka.common.serialization.Serdes$StringSerde

If this property is not set, it will use the default SerDe: spring.cloud.stream.kafka.streams.binder.configuration.default.value.serde.

It is worth to mention that Kafka Streams binder does not deserialize the keys on inbound - it simply relies on Kafka itself. Therefore, you either have to specify the keySerde property on the binding or it will default to the application-wide common keySerde.

Binding level key serde:

spring.cloud.stream.kafka.streams.bindings.input.consumer.keySerde

Common Key serde:

spring.cloud.stream.kafka.streams.binder.configuration.default.key.serde

As in the case of KStream branching on the outbound, the benefit of setting value SerDe per binding is that if you have multiple input bindings (multiple KStreams object) and they all require separate value SerDe’s, then you can configure them individually. If you use the common configuration approach, then this feature won’t be applicable.

40.7 Error Handling

Apache Kafka Streams provide the capability for natively handling exceptions from deserialization errors. For details on this support, please see this Out of the box, Apache Kafka Streams provide two kinds of deserialization exception handlers - logAndContinue and logAndFail. As the name indicates, the former will log the error and continue processing the next records and the latter will log the error and fail. LogAndFail is the default deserialization exception handler.

40.7.1 Handling Deserialization Exceptions

Kafka Streams binder supports a selection of exception handlers through the following properties.

spring.cloud.stream.kafka.streams.binder.serdeError: logAndContinue

In addition to the above two deserialization exception handlers, the binder also provides a third one for sending the erroneous records (poison pills) to a DLQ topic. Here is how you enable this DLQ exception handler.

spring.cloud.stream.kafka.streams.binder.serdeError: sendToDlq

When the above property is set, all the deserialization error records are automatically sent to the DLQ topic.

spring.cloud.stream.kafka.streams.bindings.input.consumer.dlqName: foo-dlq

If this is set, then the error records are sent to the topic foo-dlq. If this is not set, then it will create a DLQ topic with the name error.<input-topic-name>.<group-name>.

A couple of things to keep in mind when using the exception handling feature in Kafka Streams binder.

  • The property spring.cloud.stream.kafka.streams.binder.serdeError is applicable for the entire application. This implies that if there are multiple StreamListener methods in the same application, this property is applied to all of them.
  • The exception handling for deserialization works consistently with native deserialization and framework provided message conversion.

40.7.2 Handling Non-Deserialization Exceptions

For general error handling in Kafka Streams binder, it is up to the end user applications to handle application level errors. As a side effect of providing a DLQ for deserialization exception handlers, Kafka Streams binder provides a way to get access to the DLQ sending bean directly from your application. Once you get access to that bean, you can programmatically send any exception records from your application to the DLQ.

It continues to remain hard to robust error handling using the high-level DSL; Kafka Streams doesn’t natively support error handling yet.

However, when you use the low-level Processor API in your application, there are options to control this behavior. See below.

@Autowired
private SendToDlqAndContinue dlqHandler;

@StreamListener("input")
@SendTo("output")
public KStream<?, WordCount> process(KStream<Object, String> input) {

    input.process(() -> new Processor() {
    			ProcessorContext context;

    			@Override
    			public void init(ProcessorContext context) {
    				this.context = context;
    			}

    			@Override
    			public void process(Object o, Object o2) {

    			    try {
    			        .....
    			        .....
    			    }
    			    catch(Exception e) {
    			        //explicitly provide the kafka topic corresponding to the input binding as the first argument.
                        //DLQ handler will correctly map to the dlq topic from the actual incoming destination.
                        dlqHandler.sendToDlq("topic-name", (byte[]) o1, (byte[]) o2, context.partition());
    			    }
    			}

    			.....
    			.....
    });
}

40.8 State Store

State store is created automatically by Kafka Streams when the DSL is used. When processor API is used, you need to register a state store manually. In order to do so, you can use KafkaStreamsStateStore annotation. You can specify the name and type of the store, flags to control log and disabling cache, etc. Once the store is created by the binder during the bootstrapping phase, you can access this state store through the processor API. Below are some primitives for doing this.

Creating a state store:

@KafkaStreamsStateStore(name="mystate", type= KafkaStreamsStateStoreProperties.StoreType.WINDOW, lengthMs=300000)
public void process(KStream<Object, Product> input) {
    ...
}

Accessing the state store:

Processor<Object, Product>() {

    WindowStore<Object, String> state;

    @Override
    public void init(ProcessorContext processorContext) {
        state = (WindowStore)processorContext.getStateStore("mystate");
    }
    ...
}

40.9 Interactive Queries

As part of the public Kafka Streams binder API, we expose a class called InteractiveQueryService. You can access this as a Spring bean in your application. An easy way to get access to this bean from your application is to "autowire" the bean.

@Autowired
private InteractiveQueryService interactiveQueryService;

Once you gain access to this bean, then you can query for the particular state-store that you are interested. See below.

ReadOnlyKeyValueStore<Object, Object> keyValueStore =
						interactiveQueryService.getQueryableStoreType("my-store", QueryableStoreTypes.keyValueStore());

If there are multiple instances of the kafka streams application running, then before you can query them interactively, you need to identify which application instance hosts the key. InteractiveQueryService API provides methods for identifying the host information.

In order for this to work, you must configure the property application.server as below:

spring.cloud.stream.kafka.streams.binder.configuration.application.server: <server>:<port>

Here are some code snippets:

org.apache.kafka.streams.state.HostInfo hostInfo = interactiveQueryService.getHostInfo("store-name",
						key, keySerializer);

if (interactiveQueryService.getCurrentHostInfo().equals(hostInfo)) {

    //query from the store that is locally available
}
else {
    //query from the remote host
}

40.10 Accessing the underlying KafkaStreams object

StreamBuilderFactoryBean from spring-kafka that is responsible for constructing the KafkaStreams object can be accessed programmatically. Each StreamBuilderFactoryBean is registered as stream-builder and appended with the StreamListener method name. If your StreamListener method is named as process for example, the stream builder bean is named as stream-builder-process. Since this is a factory bean, it should be accessed by prepending an ampersand (&) when accessing it programmatically. Following is an example and it assumes the StreamListener method is named as process

StreamsBuilderFactoryBean streamsBuilderFactoryBean = context.getBean("&stream-builder-process", StreamsBuilderFactoryBean.class);
			KafkaStreams kafkaStreams = streamsBuilderFactoryBean.getKafkaStreams();

40.11 State Cleanup

By default, the Kafkastreams.cleanup() method is called when the binding is stopped. See the Spring Kafka documentation. To modify this behavior simply add a single CleanupConfig @Bean (configured to clean up on start, stop, or neither) to the application context; the bean will be detected and wired into the factory bean.