Kafka Streams Binder

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>

A quick way to bootstrap a new project for Kafka Streams binder is to use Spring Initializr and then select "Cloud Streams" and "Spring for Kafka Streams" as shown below

Spring Cloud Stream 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 foundations provided by the Spring for Apache Kafka project.

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

Kafka Streams applications typically follow a model in which the records are read from an inbound topic, apply business logic, and then write the transformed records to an outbound topic. Alternatively, a Processor application with no outbound destination can be defined as well.

In the following sections, we are going to look at the details of Spring Cloud Stream’s integration with Kafka Streams.

Programming Model

When using the programming model provided by Kafka Streams binder, both the high-level Streams DSL and a mix of both the higher level and the lower level Processor-API can be used as options. When mixing both higher and lower level API’s, this is usually achieved by invoking transform or process API methods on KStream.

Functional Style

Starting with Spring Cloud Stream 3.0.0, Kafka Streams binder allows the applications to be designed and developed using the functional programming style that is available in Java 8. This means that the applications can be concisely represented as a lambda expression of types java.util.function.Function or java.util.function.Consumer.

Let’s take a very basic example.

@SpringBootApplication
public class SimpleConsumerApplication {

    @Bean
    public java.util.function.Consumer<KStream<Object, String>> process() {

        return input ->
                input.foreach((key, value) -> {
                    System.out.println("Key: " + key + " Value: " + value);
                });
    }
}

Albeit simple, this is a complete standalone Spring Boot application that is leveraging Kafka Streams for stream processing. This is a consumer application with no outbound binding and only a single inbound binding. The application consumes data and it simply logs the information from the KStream key and value on the standard output. The application contains the SpringBootApplication annotation and a method that is marked as Bean. The bean method is of type java.util.function.Consumer which is parameterized with KStream. Then in the implementation, we are returning a Consumer object that is essentially a lambda expression. Inside the lambda expression, the code for processing the data is provided.

In this application, there is a single input binding that is of type KStream. The binder creates this binding for the application with a name process-in-0, i.e. the name of the function bean name followed by a dash character (-) and the literal in followed by another dash and then the ordinal position of the parameter. You use this binding name to set other properties such as destination. For example, spring.cloud.stream.bindings.process-in-0.destination=my-topic.

If the destination property is not set on the binding, a topic is created with the same name as the binding (if there are sufficient privileges for the application) or that topic is expected to be already available.

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

java -jar kstream-consumer-app.jar --spring.cloud.stream.bindings.process-in-0.destination=my-topic

Here is another example, where it is a full processor with both input and output bindings. This is the classic word-count example in which the application receives data from a topic, the number of occurrences for each word is then computed in a tumbling time-window.

@SpringBootApplication
public class WordCountProcessorApplication {

  @Bean
  public Function<KStream<Object, String>, KStream<?, WordCount>> process() {

    return input -> input
                .flatMapValues(value -> Arrays.asList(value.toLowerCase().split("\\W+")))
                .map((key, value) -> new KeyValue<>(value, value))
                .groupByKey(Serialized.with(Serdes.String(), Serdes.String()))
                .windowedBy(TimeWindows.of(5000))
                .count(Materialized.as("word-counts-state-store"))
                .toStream()
                .map((key, value) -> new KeyValue<>(key.key(), 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);
	}
}

Here again, this is a complete Spring Boot application. The difference here from the first application is that the bean method is of type java.util.function.Function. The first parameterized type for the Function is for the input KStream and the second one is for the output. In the method body, a lambda expression is provided that is of type Function and as implementation, the actual business logic is given. Similar to the previously discussed Consumer based application, the input binding here is named as process-in-0 by default. For the output, the binding name is automatically also set to process-out-0.

Once built as an 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.process-in-0.destination=words --spring.cloud.stream.bindings.process-out-0.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 Kafka Streams specific configuration required by the Kafka Streams infrastructure is automatically handled by the framework.

The two examples we saw above have a single KStream input binding. In both cases, the bindings received the records from a single topic. If you want to multiplex multiple topics into a single KStream binding, you can provide comma separated Kafka topics as destinations below.

spring.cloud.stream.bindings.process-in-0.destination=topic-1,topic-2,topic-3

In addition, you can also provide topic patterns as destinations if you want to match topics against a regular exression.

spring.cloud.stream.bindings.process-in-0.destination=input.*

Multiple Input Bindings

Many non-trivial Kafka Streams applications often consume data from more than one topic through multiple bindings. For instance, one topic is consumed as Kstream and another as KTable or GlobalKTable. There are many reasons why an application might want to receive data as a table type. Think of a use-case where the underlying topic is populated through a change data capture (CDC) mechanism from a database or perhaps the application only cares about the latest updates for downstream processing. If the application specifies that the data needs to be bound as KTable or GlobalKTable, then Kafka Streams binder will properly bind the destination to a KTable or GlobalKTable and make them available for the application to operate upon. We will look at a few different scenarios how multiple input bindings are handled in the Kafka Streams binder.

BiFunction in Kafka Streams Binder

Here is an example where we have two inputs and an output. In this case, the application can leverage on java.util.function.BiFunction.

@Bean
public BiFunction<KStream<String, Long>, KTable<String, String>, KStream<String, Long>> process() {
    return (userClicksStream, userRegionsTable) -> (userClicksStream
            .leftJoin(userRegionsTable, (clicks, region) -> new RegionWithClicks(region == null ?
                            "UNKNOWN" : region, clicks),
                    Joined.with(Serdes.String(), Serdes.Long(), null))
            .map((user, regionWithClicks) -> new KeyValue<>(regionWithClicks.getRegion(),
                    regionWithClicks.getClicks()))
            .groupByKey(Grouped.with(Serdes.String(), Serdes.Long()))
            .reduce(Long::sum)
            .toStream());
}

Here again, the basic theme is the same as in the previous examples, but here we have two inputs. Java’s BiFunction support is used to bind the inputs to the desired destinations. The default binding names generated by the binder for the inputs are process-in-0 and process-in-1 respectively. The default output binding is process-out-0. In this example, the first parameter of BiFunction is bound as a KStream for the first input and the second parameter is bound as a KTable for the second input.

BiConsumer in Kafka Streams Binder

If there are two inputs, but no outputs, in that case we can use java.util.function.BiConsumer as shown below.

@Bean
public BiConsumer<KStream<String, Long>, KTable<String, String>> process() {
    return (userClicksStream, userRegionsTable) -> {}
}

Beyond two inputs

What if you have more than two inputs? There are situations in which you need more than two inputs. In that case, the binder allows you to chain partial functions. In functional programming jargon, this technique is generally known as currying. With the functional programming support added as part of Java 8, Java now enables you to write curried functions. Spring Cloud Stream Kafka Streams binder can make use of this feature to enable multiple input bindings.

Let’s see an example.

@Bean
public Function<KStream<Long, Order>,
        Function<GlobalKTable<Long, Customer>,
                Function<GlobalKTable<Long, Product>, KStream<Long, EnrichedOrder>>>> enrichOrder() {

    return orders -> (
              customers -> (
                    products -> (
                        orders.join(customers,
                            (orderId, order) -> order.getCustomerId(),
                                (order, customer) -> new CustomerOrder(customer, order))
                                .join(products,
                                        (orderId, customerOrder) -> customerOrder
                                                .productId(),
                                        (customerOrder, product) -> {
                                            EnrichedOrder enrichedOrder = new EnrichedOrder();
                                            enrichedOrder.setProduct(product);
                                            enrichedOrder.setCustomer(customerOrder.customer);
                                            enrichedOrder.setOrder(customerOrder.order);
                                            return enrichedOrder;
                                        })
                        )
                )
    );
}

Let’s look at the details of the binding model presented above. In this model, we have 3 partially applied functions on the inbound. Let’s call them as f(x), f(y) and f(z). If we expand these functions in the sense of true mathematical functions, it will look like these: f(x) → (fy) → f(z) → KStream<Long, EnrichedOrder>. The x variable stands for KStream<Long, Order>, the y variable stands for GlobalKTable<Long, Customer> and the z variable stands for GlobalKTable<Long, Product>. The first function f(x) has the first input binding of the application (KStream<Long, Order>) and its output is the function, f(y). The function f(y) has the second input binding for the application (GlobalKTable<Long, Customer>) and its output is yet another function, f(z). The input for the function f(z) is the third input for the application (GlobalKTable<Long, Product>) and its output is KStream<Long, EnrichedOrder> which is the final output binding for the application. The input from the three partial functions which are KStream, GlobalKTable, GlobalKTable respectively are available for you in the method body for implementing the business logic as part of the lambda expression.

Input bindings are named as enrichOrder-in-0, enrichOrder-in-1 and enrichOrder-in-2 respectively. Output binding is named as enrichOrder-out-0.

With curried functions, you can virtually have any number of inputs. However, keep in mind that, anything more than a smaller number of inputs and partially applied functions for them as above in Java might lead to unreadable code. Therefore if your Kafka Streams application requires more than a reasonably smaller number of input bindings and you want to use this functional model, then you may want to rethink your design and decompose the application appropriately.

Multiple Output Bindings

Kafka Streams allows to write outbound data into multiple topics. This feature is known as branching in Kafka Streams. When using multiple output bindings, you need to provide an array of KStream (KStream[]) as the outbound return type.

Here is an example:

@Bean
public Function<KStream<Object, String>, KStream<?, WordCount>[]> process() {

    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 -> input
            .flatMapValues(value -> Arrays.asList(value.toLowerCase().split("\\W+")))
            .groupBy((key, value) -> value)
            .windowedBy(TimeWindows.of(5000))
            .count(Materialized.as("WordCounts-branch"))
            .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);
}

The programming model remains the same, however the outbound parameterized type is KStream[]. The default output binding names are process-out-0, process-out-1, process-out-2 respectively. The reason why the binder generates three output bindings is because it detects the length of the returned KStream array.

Summary of Function based Programming Styles for Kafka Streams

In summary, the following table shows the various options that can be used in the functional paradigm.

Number of Inputs	Number of Outputs	Component to use
1	0	java.util.function.Consumer
2	0	java.util.function.BiConsumer
1	1..n	java.util.function.Function
2	1..n	java.util.function.BiFunction
>= 3	0..n	Use curried functions

In the case of more than one output in this table, the type simply becomes KStream[].

Imperative programming model.

Although the functional programming model outlined above is the preferred approach, you can still use the classic StreamListener based approach if you prefer.

Here are some examples.

Following is the equivalent of the Word count example using StreamListener.

@SpringBootApplication
@EnableBinding(KafkaStreamsProcessor.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);
    }

As you can see, this is a bit more verbose since you need to provide EnableBinding and the other extra annotations like StreamListener and SendTo to make it a complete application. EnableBinding is where you specify your binding interface that contains your bindings. In this case, we are using the stock KafkaStreamsProcessor binding interface that has the following contracts.

public interface KafkaStreamsProcessor {

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

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

}

Binder will create bindings for the input KStream and output KStream since you are using a binding interface that contains those declarations.

In addition to the obvious differences in the programming model offered in the functional style, one particular thing that needs to be mentioned here is that the binding names are what you specify in the binding interface. For example, in the above application, since we are using KafkaStreamsProcessor, the binding names are input and output. Binding properties need to use those names. For instance spring.cloud.stream.bindings.input.destination, spring.cloud.stream.bindings.output.destination etc. Keep in mind that this is fundamentally different from the functional style since there the binder generates binding names for the application. This is because the application does not provide any binding interfaces in the functional model using EnableBinding.

Here is another example of a sink where we have two inputs.

@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();
}

Following is the StreamListener equivalent of the same BiFunction based processor that we saw above.

@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();
}

Finally, here is the StreamListener equivalent of the application with three inputs and curried functions.

@EnableBinding(CustomGlobalKTableProcessor.class)
...
...
    @StreamListener
    @SendTo("output")
    public KStream<Long, EnrichedOrder> process(
            @Input("input-1") KStream<Long, Order> ordersStream,
            @Input("input-2") GlobalKTable<Long, Customer> customers,
            @Input("input-3") GlobalKTable<Long, Product> products) {

        KStream<Long, CustomerOrder> customerOrdersStream = ordersStream.join(
                customers, (orderId, order) -> order.getCustomerId(),
                (order, customer) -> new CustomerOrder(customer, order));

        return customerOrdersStream.join(products,
                (orderId, customerOrder) -> customerOrder.productId(),
                (customerOrder, product) -> {
                    EnrichedOrder enrichedOrder = new EnrichedOrder();
                    enrichedOrder.setProduct(product);
                    enrichedOrder.setCustomer(customerOrder.customer);
                    enrichedOrder.setOrder(customerOrder.order);
                    return enrichedOrder;
                });
        }

    interface CustomGlobalKTableProcessor {

            @Input("input-1")
            KStream<?, ?> input1();

            @Input("input-2")
            GlobalKTable<?, ?> input2();

            @Input("input-3")
            GlobalKTable<?, ?> input3();

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

You might notice that the above two examples are even more verbose since in addition to provide EnableBinding, you also need to write your own custom binding interface as well. Using the functional model, you can avoid all those ceremonial details.

Before we move on from looking at the general programming model offered by Kafka Streams binder, here is the StreamListener version of multiple output bindings.

EnableBinding(KStreamProcessorWithBranches.class)
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();
    	}
}

To recap, we have reviewed the various programming model choices when using the Kafka Streams binder.

The binder provides binding capabilities for KStream, KTable and GlobalKTable on the input. KTable and GlobalKTable bindings are only available on the input. Binder supports both input and output bindings for KStream.

The upshot of the programming model of Kafka Streams binder is that the binder provides you the flexibility of going with a fully functional programming model or using the StreamListener based imperative approach.

Ancillaries to the programming model

Multiple Kafka Streams processors within a single application

Binder allows to have multiple Kafka Streams processors within a single Spring Cloud Stream application. You can have an application as below.

@Bean
public java.util.function.Function<KStream<Object, String>, KStream<Object, String>> process() {
   ...
}

@Bean
public java.util.function.Consumer<KStream<Object, String>> anotherProcess() {
  ...
}

@Bean
public java.util.function.BiFunction<KStream<Object, String>, KTable<Integer, String>, KStream<Object, String>> yetAnotherProcess() {
   ...
}

In this case, the binder will create 3 separate Kafka Streams objects with different application ID’s (more on this below). However, if you have more than one processor in the application, you have to tell Spring Cloud Stream, which functions need to be activated. Here is how you activate the functions.

spring.cloud.stream.function.definition: process;anotherProcess;yetAnotherProcess

If you want certain functions to be not activated right away, you can remove that from this list.

This is also true when you have a single Kafka Streams processor and other types of Function beans in the same application that is handled through a different binder (for e.g., a function bean that is based on the regular Kafka Message Channel binder)

Kafka Streams Application ID

Application id is a mandatory property that you need to provide for a Kafka Streams application. Spring Cloud Stream Kafka Streams binder allows you to configure this application id in multiple ways.

If you only have one single processor or StreamListener in the application, then you can set this at the binder level using the following property:

spring.cloud.stream.kafka.streams.binder.applicationId.

As a convenience, if you only have a single processor, you can also use spring.application.name as the property to delegate the application id.

If you have multiple Kafka Streams processors in the application, then you need to set the application id per processor. In the case of the functional model, you can attach it to each function as a property.

For e.g. imagine that you have the following functions.

@Bean
public java.util.function.Consumer<KStream<Object, String>> process() {
   ...
}

and

@Bean
public java.util.function.Consumer<KStream<Object, String>> anotherProcess() {
  ...
}

Then you can set the application id for each, using the following binder level properties.

spring.cloud.stream.kafka.streams.binder.functions.process.applicationId

and

spring.cloud.stream.kafka.streams.binder.functions.anotherProcess.applicationId

In the case of StreamListener, you need to set this on the first input binding on the processor.

For e.g. imagine that you have the following two StreamListener based processors.

@StreamListener
@SendTo("output")
public KStream<String, String> process(@Input("input") <KStream<Object, String>> input) {
   ...
}

@StreamListener
@SendTo("anotherOutput")
public KStream<String, String> anotherProcess(@Input("anotherInput") <KStream<Object, String>> input) {
   ...
}

Then you must set the application id for this using the following binding property.

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

and

spring.cloud.stream.kafka.streams.bindings.anotherInput.consumer.applicationId

For function based model also, this approach of setting application id at the binding level will work. However, setting per function at the binder level as we have seen above is much easier if you are using the functional model.

For production deployments, it is highly recommended to explicitly specify the application ID through configuration. This is especially going to be very critical if you are auto scaling your application in which case you need to make sure that you are deploying each instance with the same application ID.

If the application does not provide an application ID, then in that case the binder will auto generate a static application ID for you. This is convenient in development scenarios as it avoids the need for explicitly providing the application ID. The generated application ID in this manner will be static over application restarts. In the case of functional model, the generated application ID will be the function bean name followed by the literal applicationID, for e.g process-applicationID if process if the function bean name. In the case of StreamListener, instead of using the function bean name, the generated application ID will be use the containing class name followed by the method name followed by the literal applicationId.

Summary of setting Application ID

By default, binder will auto generate the application ID per function or StreamListener methods.
If you have a single processor, then you can use spring.kafka.streams.applicationId, spring.application.name or spring.cloud.stream.kafka.streams.binder.applicationId.
If you have multiple processors, then application ID can be set per function using the property - spring.cloud.stream.kafka.streams.binder.functions.<function-name>.applicationId. In the case of StreamListener, this can be done using spring.cloud.stream.kafka.streams.bindings.input.applicationId, assuming that the input binding name is input.

Overriding the default binding names generated by the binder with the functional style

By default, the binder uses the strategy discussed above to generate the binding name when using the functional style, i.e. <function-bean-name>-<in>|<out>-[0..n], for e.g. process-in-0, process-out-0 etc. If you want to override those binding names, you can do that by specifying the following properties.

spring.cloud.stream.function.bindings.<default binding name>. Default binding name is the original binding name generated by the binder.

For e.g. lets say, you have this function.

@Bean
public BiFunction<KStream<String, Long>, KTable<String, String>, KStream<String, Long>> process() {
...
}

Binder will generate bindings with names, process-in-0, process-in-1 and process-out-0. Now, if you want to change them to something else completely, maybe more domain specific binding names, then you can do so as below.

spring.cloud.stream.function.bindings.process-in-0=users

spring.cloud.stream.function.bindings.process-in-0=regions

and

spring.cloud.stream.function.bindings.process-out-0=clicks

After that, you must set all the binding level properties on these new binding names.

Please keep in mind that with the functional programming model described above, adhering to the default binding names make sense in most situations. The only reason you may still want to do this overriding is when you have larger number of configuration properties and you want to map the bindings to something more domain friendly.

Setting up bootstrap server configuration

When running Kafka Streams applications, you must provide the Kafka broker server information. If you don’t provide this information, the binder expects that you are running the broker at the default localhost:9092. If that is not the case, then you need to override that. There are a couple of ways to do that.

Using the boot property - spring.kafka.bootstrapServers
Binder level property - spring.cloud.stream.kafka.streams.binder.brokers

When it comes to the binder level property, it doesn’t matter if you use the broker property provided through the regular Kafka binder - spring.cloud.stream.kafka.binder.brokers. Kafka Streams binder will first check if Kafka Streams binder specific broker property is set (spring.cloud.stream.kafka.streams.binder.brokers) and if not found, it looks for spring.cloud.stream.kafka.binder.brokers.

Record serialization and deserialization

Kafka Streams binder allows you to serialize and deserialize records in two ways. One is the native serialization and deserialization facilities provided by Kafka and the other one is the message conversion capabilities of Spring Cloud Stream framework. Lets look at some details.

Inbound deserialization

Keys are always deserialized using native Serdes.

For values, by default, deserialization on the inbound is natively performed by Kafka. Please note that this is a major change on default behavior from previous versions of Kafka Streams binder where the deserialization was done by the framework.

Kafka Streams binder will try to infer matching Serde types by looking at the type signature of java.util.function.Function|Consumer or StreamListener. Here is the order that it matches the Serdes.

If the application provides a bean of type Serde and if the return type is parameterized with the actual type of the incoming key or value type, then it will use that Serde for inbound deserialization. For e.g. if you have the following in the application, the binder detects that the incoming value type for the KStream matches with a type that is parameterized on a Serde bean. It will use that for inbound deserialization.

@Bean
public Serde<Foo() customSerde{
 ...
}

@Bean
public Function<KStream<String, Foo>, KStream<String, Foo>> process() {
}

Next, it looks at the types and see if they are one of the types exposed by Kafka Streams. If so, use them. Here are the Serde types that the binder will try to match from Kafka Streams.
```
Integer, Long, Short, Double, Float, byte[], UUID and String.
```
If none of the Serdes provided by Kafka Streams don’t match the types, then it will use JsonSerde provided by Spring Kafka. In this case, the binder assumes that the types are JSON friendly. This is useful if you have multiple value objects as inputs since the binder will internally infer them to correct Java types. Before falling back to the JsonSerde though, the binder checks at the default Serde`s set in the Kafka Streams configuration to see if it is a `Serde that it can match with the incoming KStream’s types.

If none of the above strategies worked, then the applications must provide the `Serde`s through configuration. This can be configured in two ways - binding or default.

First the binder will look if a Serde is provided at the binding level. For e.g. if you have the following processor,

@Bean
public BiFunction<KStream<CustomKey, AvroIn1>, KTable<CustomKey, AvroIn2>, KStream<CustomKey, AvroOutput>> process() {...}

then, you can provide a binding level Serde using the following:

spring.cloud.stream.kafka.streams.bindings.process-in-0.consumer.keySerde=CustomKeySerde
spring.cloud.stream.kafka.streams.bindings.process-in-0.consumer.valueSerde=io.confluent.kafka.streams.serdes.avro.SpecificAvroSerde

spring.cloud.stream.kafka.streams.bindings.process-in-1.consumer.keySerde=CustomKeySerde
spring.cloud.stream.kafka.streams.bindings.process-in-1.consumer.valueSerde=io.confluent.kafka.streams.serdes.avro.SpecificAvroSerde

If you provide Serde as abover per input binding, then that will takes higher precedence and the binder will stay away from any Serde inference.

If you want the default key/value Serdes to be used for inbound deserialization, you can do so at the binder level.

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

If you don’t want the native decoding provided by Kafka, you can rely on the message conversion features that Spring Cloud Stream provides. Since native decoding is the default, in order to let Spring Cloud Stream deserialize the inbound value object, you need to explicitly disable native decoding.

For e.g. if you have the same BiFunction processor as above, then spring.cloud.stream.bindings.process-in-0.consumer.nativeDecoding: false You need to disable native decoding for all the inputs individually. Otherwise, native decoding will still be applied for those you do not disable.

By default, Spring Cloud Stream will use application/json as the content type and use an appropriate json message converter. You can use custom message converters by using the following property and an appropriate MessageConverter bean.

spring.cloud.stream.bindings.process-in-0.contentType

Outbound serialization

Outbound serialization pretty much follows the same rules as above for inbound deserialization. As with the inbound deserialization, one major change from the previous versions of Spring Cloud Stream is that the serialization on the outbound is handled by Kafka natively. Before 3.0 versions of the binder, this was done by the framework itself.

Keys on the outbound are always serialized by Kafka using a matching Serde that is inferred by the binder. If it can’t infer the type of the key, then that needs to be specified using configuration.

Value serdes are inferred using the same rules used for inbound deserialization. First it matches to see if the outbound type is from a provided bean in the application. If not, it checks to see if it matches with a Serde exposed by Kafka such as - Integer, Long, Short, Double, Float, byte[], UUID and String. If that doesnt’t work, then it falls back to JsonSerde provided by the Spring Kafka project, but first look at the default Serde configuration to see if there is a match. Keep in mind that all these happen transparently to the application. If none of these work, then the user has to provide the Serde to use by configuration.

Lets say you are using the same BiFunction processor as above. Then you can configure outbound key/value Serdes as following.

spring.cloud.stream.kafka.streams.bindings.process-out-0.producer.keySerde=CustomKeySerde
spring.cloud.stream.kafka.streams.bindings.process-out-0.producer.valueSerde=io.confluent.kafka.streams.serdes.avro.SpecificAvroSerde

If Serde inference fails, and no binding level Serdes are provided, then the binder falls back to the JsonSerde, but look at the default Serdes for a match.

Default serdes are configured in the same way as above where it is described under deserialization.

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

If your application uses the branching feature and has multiple output bindings, then these have to be configured per binding. Once again, if the binder is capable of inferring the Serde types, you don’t need to do this configuration.

If you don’t want the native encoding provided by Kafka, but want to use the framework provided message conversion, then you need to explicitly disable native encoding since since native encoding is the default. For e.g. if you have the same BiFunction processor as above, then spring.cloud.stream.bindings.process-out-0.producer.nativeEncoding: false You need to disable native encoding for all the output individually in the case of branching. Otherwise, native encoding will still be applied for those you don’t disable.

When conversion is done by Spring Cloud Stream, by default, it will use application/json as the content type and use an appropriate json message converter. You can use custom message converters by using the following property and a corresponding MessageConverter bean.

spring.cloud.stream.bindings.process-out-0.contentType

When native encoding/decoding is disabled, binder will not do any inference as in the case of native Serdes. Applications need to explicitly provide all the configuration options. For that reason, it is generally advised to stay with the default options for de/serialization and stick with native de/serialization provided by Kafka Streams when you write Spring Cloud Stream Kafka Streams applications. The one scenario in which you must use message conversion capabilities provided by the framework is when your upstream producer is using a specific serialization strategy. In that case, you want to use a matching deserialization strategy as native mechanisms may fail. When relying on the default Serde mechanism, the applications must ensure that the binder has a way forward with correctly map the inbound and outbound with a proper Serde, as otherwise things might fail.

It is worth to mention that the data de/serialization approaches outlined above are only applicable on the edges of your processors, i.e. - inbound and outbound. Your business logic might still need to call Kafka Streams API’s that explicitly need Serde objects. Those are still the responsibility of the application and must be handled accordingly by the developer.

Error Handling

Apache Kafka Streams provides the capability for natively handling exceptions from deserialization errors. For details on this support, please see this. Out of the box, Apache Kafka Streams provides two kinds of deserialization exception handlers - LogAndContinueExceptionHandler and LogAndFailExceptionHandler. 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. LogAndFailExceptionHandler is the default deserialization exception handler.

Handling Deserialization Exceptions in the Binder

Kafka Streams binder allows to specify the deserialization exception handlers above using the following property.

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

spring.cloud.stream.kafka.streams.binder.deserializationExceptionHandler: logAndFail

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 (dead letter queue) topic. Here is how you enable this DLQ exception handler.

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

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

You can set the topic name where the DLQ messages are published as below.

You can provide an implementation for DlqDestinationResolver which is a functional interface. DlqDestinationResolver takes ConsumerRecord and the exception as inputs and then allows to specify a topic name as the output. By gaining access to the Kafka ConsumerRecord, the header records can be introspected in the implementation of the BiFunction.

Here is an example of providing an implementation for DlqDestinationResolver.

@Bean
public DlqDestinationResolver dlqDestinationResolver() {
    return (rec, ex) -> {
        if (rec.topic().equals("word1")) {
            return "topic1-dlq";
        }
        else {
            return "topic2-dlq";
        }
    };
}

One important thing to keep in mind when providing an implementation for DlqDestinationResolver is that the provisioner in the binder will not auto create topics for the application. This is because there is no way for the binder to infer the names of all the DLQ topics the implementation might send to. Therefore, if you provide DLQ names using this strategy, it is the application’s responsibility to ensure that those topics are created beforehand.

If DlqDestinationResolver is present in the application as a bean, that takes higher prcedence. If you do not want to follow this approach and rather provide a static DLQ name using configuration, you can set the following property.

spring.cloud.stream.kafka.streams.bindings.process-in-0.consumer.dlqName: custom-dlq (Change the binding name accordingly)

If this is set, then the error records are sent to the topic custom-dlq. If the application is not using either of the above strategies, then it will create a DLQ topic with the name error.<input-topic-name>.<application-id>. For instance, if your binding’s destination topic is inputTopic and the application ID is process-applicationId, then the default DLQ topic is error.inputTopic.process-applicationId. It is always recommended to explicitly create a DLQ topic for each input binding if it is your intention to enable DLQ.

DLQ per input consumer binding

The property spring.cloud.stream.kafka.streams.binder.deserializationExceptionHandler is applicable for the entire application. This implies that if there are multiple functions or StreamListener methods in the same application, this property is applied to all of them. However, if you have multiple processors or multiple input bindings within a single processor, then you can use the finer-grained DLQ control that the binder provides per input consumer binding.

If you have the following processor,

@Bean
public BiFunction<KStream<String, Long>, KTable<String, String>, KStream<String, Long>> process() {
...
}

and you only want to enable DLQ on the first input binding and logAndSkip on the second binding, then you can do so on the consumer as below.

spring.cloud.stream.kafka.streams.bindings.process-in-0.consumer.deserializationExceptionHandler: sendToDlq spring.cloud.stream.kafka.streams.bindings.process-in-1.consumer.deserializationExceptionHandler: logAndSkip

Setting deserialization exception handlers this way has a higher precedence than setting at the binder level.

DLQ partitioning

By default, records are published to the Dead-Letter topic using the same partition as the original record. This means the Dead-Letter topic must have at least as many partitions as the original record.

To change this behavior, add a DlqPartitionFunction implementation as a @Bean to the application context. Only one such bean can be present. The function is provided with the consumer group (which is the same as the application ID in most situations), the failed ConsumerRecord and the exception. For example, if you always want to route to partition 0, you might use:

@Bean
public DlqPartitionFunction partitionFunction() {
    return (group, record, ex) -> 0;
}

If you set a consumer binding’s dlqPartitions property to 1 (and the binder’s minPartitionCount is equal to 1), there is no need to supply a DlqPartitionFunction; the framework will always use partition 0. If you set a consumer binding’s dlqPartitions property to a value greater than 1 (or the binder’s minPartitionCount is greater than 1), you must provide a DlqPartitionFunction bean, even if the partition count is the same as the original topic’s.

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.deserializationExceptionHandler is applicable for the entire application. This implies that if there are multiple functions or 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.

Handling Production Exceptions in the Binder

Unlike the support for deserialization exception handlers as described above, the binder does not provide such first class mechanisms for handling production exceptions. However, you still can configure production exception handlers using the StreamsBuilderFactoryBean customizer which you can find more details about, in a subsequent section below.

Retrying critical business logic

There are scenarios in which you might want to retry parts of your business logic that are critical to the application. There maybe an external call to a relational database or invoking a REST endpoint from the Kafka Streams processor. These calls can fail for various reasons such as network issues or remote service unavailability. More often, these failures may self resolve if you can try them again. By default, Kafka Streams binder creates RetryTemplate beans for all the input bindings.

If the function has the following signature,

@Bean
public java.util.function.Consumer<KStream<Object, String>> process()

and with default binding name, the RetryTemplate will be registered as process-in-0-RetryTemplate. This is following the convention of binding name (process-in-0) followed by the literal -RetryTemplate. In the case of multiple input bindings, there will be a separate RetryTemplate bean available per binding. If there is a custom RetryTemplate bean available in the application and provided through spring.cloud.stream.bindings.<binding-name>.consumer.retryTemplateName, then that takes precedence over any input binding level retry template configuration properties.

Once the RetryTemplate from the binding is injected into the application, it can be used to retry any critical sections of the application. Here is an example:

@Bean
public java.util.function.Consumer<KStream<Object, String>> process(@Lazy @Qualifier("process-in-0-RetryTemplate") RetryTemplate retryTemplate) {

    return input -> input
            .process(() -> new Processor<Object, String>() {
                @Override
                public void init(ProcessorContext processorContext) {
                }

                @Override
                public void process(Object o, String s) {
                    retryTemplate.execute(context -> {
                       //Critical business logic goes here.
                    });
                }

                @Override
                public void close() {
                }
            });
}

Or you can use a custom RetryTemplate as below.

@EnableAutoConfiguration
public static class CustomRetryTemplateApp {

    @Bean
    @StreamRetryTemplate
    RetryTemplate fooRetryTemplate() {
        RetryTemplate retryTemplate = new RetryTemplate();

        RetryPolicy retryPolicy = new SimpleRetryPolicy(4);
        FixedBackOffPolicy backOffPolicy = new FixedBackOffPolicy();
        backOffPolicy.setBackOffPeriod(1);

        retryTemplate.setBackOffPolicy(backOffPolicy);
        retryTemplate.setRetryPolicy(retryPolicy);

        return retryTemplate;
    }

    @Bean
    public java.util.function.Consumer<KStream<Object, String>> process() {

        return input -> input
                .process(() -> new Processor<Object, String>() {
                    @Override
                    public void init(ProcessorContext processorContext) {
                    }

                    @Override
                    public void process(Object o, String s) {
                        fooRetryTemplate().execute(context -> {
                           //Critical business logic goes here.
                        });

                    }

                    @Override
                    public void close() {
                    }
                });
    }
}

Note that when retries are exhausted, by default, the last exception will be thrown, causing the processor to terminate. If you wish to handle the exception and continue processing, you can add a RecoveryCallback to the execute method: Here is an example.

retryTemplate.execute(context -> {
    //Critical business logic goes here.
    }, context -> {
       //Recovery logic goes here.
       return null;
    ));

Refer to the Spring Retry project for more information about the RetryTemplate, retry policies, backoff policies and more.

State Store

State stores are created automatically by Kafka Streams when the high level DSL is used and appropriate calls are made those trigger a state store.

If you want to materialize an incoming KTable binding as a named state store, then you can do so by using the following strategy.

Lets say you have the following function.

@Bean
public BiFunction<KStream<String, Long>, KTable<String, String>, KStream<String, Long>> process() {
   ...
}

Then by setting the following property, the incoming KTable data will be materialized in to the named state store.

spring.cloud.stream.kafka.streams.bindings.process-in-1.consumer.materializedAs: incoming-store

You can define custom state stores as beans in your application and those will be detected and added to the Kafka Streams builder by the binder. Especially when the processor API is used, you need to register a state store manually. In order to do so, you can create the StateStore as a bean in the application. Here are examples of defining such beans.

@Bean
public StoreBuilder myStore() {
    return Stores.keyValueStoreBuilder(
            Stores.persistentKeyValueStore("my-store"), Serdes.Long(),
            Serdes.Long());
}

@Bean
public StoreBuilder otherStore() {
    return Stores.windowStoreBuilder(
            Stores.persistentWindowStore("other-store",
                    1L, 3, 3L, false), Serdes.Long(),
            Serdes.Long());
}

These state stores can be then accessed by the applications directly.

During the bootstrap, the above beans will be processed by the binder and passed on to the Streams builder object.

Accessing the state store:

Processor<Object, Product>() {

    WindowStore<Object, String> state;

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

This will not work when it comes to registering global state stores. In order to register a global state store, please see the section below on customizing StreamsBuilderFactoryBean.

Interactive Queries

Kafka Streams binder API exposes a class called InteractiveQueryService to interactively query the state stores. 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());

During the startup, the above method call to retrieve the store might fail. For e.g it might still be in the middle of initializing the state store. In such cases, it will be useful to retry this operation. Kafka Streams binder provides a simple retry mechanism to accommodate this.

Following are the two properties that you can use to control this retrying.

spring.cloud.stream.kafka.streams.binder.stateStoreRetry.maxAttempts - Default is 1 .
spring.cloud.stream.kafka.streams.binder.stateStoreRetry.backOffInterval - Default is 1000 milliseconds.

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 particular key that you are querying. 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
}

Other API methods available through the InteractiveQueryService

Use the following API method to retrieve the KeyQueryMetadata object associated with the combination of given store and key.

public <K> KeyQueryMetadata getKeyQueryMetadata(String store, K key, Serializer<K> serializer)

Use the following API method to retrieve the KakfaStreams object associated with the combination of given store and key.

public <K> KafkaStreams getKafkaStreams(String store, K key, Serializer<K> serializer)

Health Indicator

The health indicator requires the dependency spring-boot-starter-actuator. For maven use:

<dependency>
  <groupId>org.springframework.boot</groupId>
  <artifactId>spring-boot-starter-actuator</artifactId>
</dependency>

Spring Cloud Stream Kafka Streams Binder provides a health indicator to check the state of the underlying streams threads. Spring Cloud Stream defines a property management.health.binders.enabled to enable the health indicator. See the Spring Cloud Stream documentation.

The health indicator provides the following details for each stream thread’s metadata:

Thread name
Thread state: CREATED, RUNNING, PARTITIONS_REVOKED, PARTITIONS_ASSIGNED, PENDING_SHUTDOWN or DEAD
Active tasks: task ID and partitions
Standby tasks: task ID and partitions

By default, only the global status is visible (UP or DOWN). To show the details, the property management.endpoint.health.show-details must be set to ALWAYS or WHEN_AUTHORIZED. For more details about the health information, see the Spring Boot Actuator documentation.

The status of the health indicator is UP if all the Kafka threads registered are in the RUNNING state.

Since there are three individual binders in Kafka Streams binder (KStream, KTable and GlobalKTable), all of them will report the health status. When enabling show-details, some of the information reported may be redundant.

When there are multiple Kafka Streams processors present in the same application, then the health checks will be reported for all of them and will be categorized by the application ID of Kafka Streams.

Accessing Kafka Streams Metrics

Spring Cloud Stream Kafka Streams binder provides Kafka Streams metrics which can be exported through a Micrometer MeterRegistry.

For Spring Boot version 2.2.x, the metrics support is provided through a custom Micrometer metrics implementation by the binder. For Spring Boot version 2.3.x, the Kafka Streams metrics support is provided natively through Micrometer.

When accessing metrics through the Boot actuator endpoint, make sure to add metrics to the property management.endpoints.web.exposure.include. Then you can access /acutator/metrics to get a list of all the available metrics, which then can be individually accessed through the same URI (/actuator/metrics/<metric-name>).

Mixing high level DSL and low level Processor API

Kafka Streams provides two variants of APIs. It has a higher level DSL like API where you can chain various operations that maybe familiar to a lot of functional programmers. Kafka Streams also gives access to a low level Processor API. The processor API, although very powerful and gives the ability to control things in a much lower level, is imperative in nature. Kafka Streams binder for Spring Cloud Stream, allows you to use either the high level DSL or mixing both the DSL and the processor API. Mixing both of these variants give you a lot of options to control various use cases in an application. Applications can use the transform or process method API calls to get access to the processor API.

Here is a look at how one may combine both the DSL and the processor API in a Spring Cloud Stream application using the process API.

@Bean
public Consumer<KStream<Object, String>> process() {
    return input ->
        input.process(() -> new Processor<Object, String>() {
            @Override
            @SuppressWarnings("unchecked")
            public void init(ProcessorContext context) {
               this.context = context;
            }

            @Override
            public void process(Object key, String value) {
                //business logic
            }

            @Override
            public void close() {

        });
}

Here is an example using the transform API.

@Bean
public Consumer<KStream<Object, String>> process() {
    return (input, a) ->
        input.transform(() -> new Transformer<Object, String, KeyValue<Object, String>>() {
            @Override
            public void init(ProcessorContext context) {

            }

            @Override
            public void close() {

            }

            @Override
            public KeyValue<Object, String> transform(Object key, String value) {
                // business logic - return transformed KStream;
            }
        });
}

The process API method call is a terminal operation while the transform API is non terminal and gives you a potentially transformed KStream using which you can continue further processing using either the DSL or the processor API.

Partition support on the outbound

A Kafka Streams processor usually sends the processed output into an outbound Kafka topic. If the outbound topic is partitioned and the processor needs to send the outgoing data into particular partitions, the applications needs to provide a bean of type StreamPartitioner. See StreamPartitioner for more details. Let’s see some examples.

This is the same processor we already saw multiple times,

@Bean
public Function<KStream<Object, String>, KStream<?, WordCount>> process() {

    ...
}

Here is the output binding destination:

spring.cloud.stream.bindings.process-out-0.destination: outputTopic

If the topic outputTopic has 4 partitions, if you don’t provide a partitioning strategy, Kafka Streams will use default partitioning strategy which may not be the outcome you want depending on the particular use case. Let’s say, you want to send any key that matches to spring to partition 0, cloud to partition 1, stream to partition 2, and everything else to partition 3. This is what you need to do in the application.

@Bean
public StreamPartitioner<String, WordCount> streamPartitioner() {
    return (t, k, v, n) -> {
        if (k.equals("spring")) {
            return 0;
        }
        else if (k.equals("cloud")) {
            return 1;
        }
        else if (k.equals("stream")) {
            return 2;
        }
        else {
            return 3;
        }
    };
}

This is a rudimentary implementation, however, you have access to the key/value of the record, the topic name and the total number of partitions. Therefore, you can implement complex partitioning strategies if need be.

You also need to provide this bean name along with the application configuration.

spring.cloud.stream.kafka.streams.bindings.process-out-0.producer.streamPartitionerBeanName: streamPartitioner

Each output topic in the application needs to be configured separately like this.

StreamsBuilderFactoryBean customizer

It is often required to customize the StreamsBuilderFactoryBean that creates the KafkaStreams objects. Based on the underlying support provided by Spring Kafka, the binder allows you to customize the StreamsBuilderFactoryBean. You can use the StreamsBuilderFactoryBeanCustomizer to customize the StreamsBuilderFactoryBean itself. Then, once you get access to the StreamsBuilderFactoryBean through this customizer, you can customize the corresponding KafkaStreams using KafkaStreamsCustomzier. Both of these customizers are part of the Spring for Apache Kafka project.

Here is an example of using the StreamsBuilderFactoryBeanCustomizer.

@Bean
public StreamsBuilderFactoryBeanCustomizer streamsBuilderFactoryBeanCustomizer() {
    return sfb -> sfb.setStateListener((newState, oldState) -> {
         //Do some action here!
    });
}

The above is shown as an illustration of the things you can do to customize the StreamsBuilderFactoryBean. You can essentially call any available mutation operations from StreamsBuilderFactoryBean to customize it. This customizer will be invoked by the binder right before the factory bean is started.

Once you get access to the StreamsBuilderFactoryBean, you can also customize the underlying KafkaStreams object. Here is a blueprint for doing so.

@Bean
public StreamsBuilderFactoryBeanCustomizer streamsBuilderFactoryBeanCustomizer() {
    return factoryBean -> {
        factoryBean.setKafkaStreamsCustomizer(new KafkaStreamsCustomizer() {
            @Override
            public void customize(KafkaStreams kafkaStreams) {
                kafkaStreams.setUncaughtExceptionHandler((t, e) -> {

                });
            }
        });
    };
}

KafkaStreamsCustomizer will be called by the StreamsBuilderFactoryBeabn right before the underlying KafkaStreams gets started.

There can only be one StreamsBuilderFactoryBeanCustomizer in the entire application. Then how do we account for multiple Kafka Streams processors as each of them are backed up by individual StreamsBuilderFactoryBean objects? In that case, if the customization needs to be different for those processors, then the application needs to apply some filter based on the application ID.

For e.g,

@Bean
public StreamsBuilderFactoryBeanCustomizer streamsBuilderFactoryBeanCustomizer() {

    return factoryBean -> {
        if (factoryBean.getStreamsConfiguration().getProperty(StreamsConfig.APPLICATION_ID_CONFIG)
                .equals("processor1-application-id")) {
            factoryBean.setKafkaStreamsCustomizer(new KafkaStreamsCustomizer() {
                @Override
                public void customize(KafkaStreams kafkaStreams) {
                    kafkaStreams.setUncaughtExceptionHandler((t, e) -> {

                    });
                }
            });
        }
    };

Using Customizer to register a global state store

As mentioned above, the binder does not provide a first class way to register global state stores as a feature. For that, you need to use the customizer. Here is how that can be done.

@Bean
public StreamsBuilderFactoryBeanCustomizer customizer() {
    return fb -> {
        try {
            final StreamsBuilder streamsBuilder = fb.getObject();
            streamsBuilder.addGlobalStore(...);
        }
        catch (Exception e) {

        }
    };
}

Again, if you have multiple processors, you want to attach the global state store to the right StreamsBuilder by filtering out the other StreamsBuilderFactoryBean objects using the application id as outlined above.

Using customizer to register a production exception handler

In the error handling section, we indicated that the binder does not provide a first class way to deal with production exceptions. Though that is the case, you can still use the StreamsBuilderFacotryBean customizer to register production exception handlers. See below.

@Bean
public StreamsBuilderFactoryBeanCustomizer customizer() {
    return fb -> {
        fb.getStreamsConfiguration().put(StreamsConfig.DEFAULT_PRODUCTION_EXCEPTION_HANDLER_CLASS_CONFIG,
                            CustomProductionExceptionHandler.class);
    };
}

Once again, if you have multiple processors, you may want to set it appropriately against the correct StreamsBuilderFactoryBean. You may also add such production exception handlers using the configuration property (See below for more on that), but this is an option if you choose to go with a programmatic approach.

Timestamp extractor

Kafka Streams allows you to control the processing of the consumer records based on various notions of timestamp. By default, Kafka Streams extracts the timestamp metadata embedded in the consumer record. You can change this default behavior by providing a different TimestampExtractor implementation per input binding. Here are some details on how that can be done.

@Bean
public Function<KStream<Long, Order>,
        Function<KTable<Long, Customer>,
                Function<GlobalKTable<Long, Product>, KStream<Long, Order>>>> process() {
    return orderStream ->
            customers ->
                products -> orderStream;
}

@Bean
public TimestampExtractor timestampExtractor() {
    return new WallclockTimestampExtractor();
}

Then you set the above TimestampExtractor bean name per consumer binding.

spring.cloud.stream.kafka.streams.bindings.process-in-0.consumer.timestampExtractorBeanName=timestampExtractor
spring.cloud.stream.kafka.streams.bindings.process-in-1.consumer.timestampExtractorBeanName=timestampExtractor
spring.cloud.stream.kafka.streams.bindings.process-in-2.consumer.timestampExtractorBeanName=timestampExtractor"

If you skip an input consumer binding for setting a custom timestamp extractor, that consumer will use the default settings.

Multi binders with Kafka Streams based binders and regular Kafka Binder

You can have an application where you have both a function/consumer/supplier that is based on the regular Kafka binder and a Kafka Streams based processor. However, you cannot mix both of them within a single function or consumer.

Here is an example, where you have both binder based components within the same application.

@Bean
public Function<String, String> process() {
    return s -> s;
}

@Bean
public Function<KStream<Object, String>, KStream<?, WordCount>> kstreamProcess() {

    return input -> input;
}

This is the relevant parts from the configuration:

spring.cloud.stream.function.definition=process;kstreamProcess
spring.cloud.stream.bindings.process-in-0.destination=foo
spring.cloud.stream.bindings.process-out-0.destination=bar
spring.cloud.stream.bindings.kstreamProcess-in-0.destination=bar
spring.cloud.stream.bindings.kstreamProcess-out-0.destination=foobar

Things become a bit more complex if you have the same application as above, but is dealing with two different Kafka clusters, for e.g. the regular process is acting upon both Kafka cluster 1 and cluster 2 (receiving data from cluster-1 and sending to cluster-2) and the Kafka Streams processor is acting upon Kafka cluster 2. Then you have to use the multi binder facilities provided by Spring Cloud Stream.

Here is how your configuration may change in that scenario.

# multi binder configuration
spring.cloud.stream.binders.kafka1.type: kafka
spring.cloud.stream.binders.kafka1.environment.spring.cloud.stream.kafka.streams.binder.brokers=${kafkaCluster-1} #Replace kafkaCluster-1 with the approprate IP of the cluster
spring.cloud.stream.binders.kafka2.type: kafka
spring.cloud.stream.binders.kafka2.environment.spring.cloud.stream.kafka.streams.binder.brokers=${kafkaCluster-2} #Replace kafkaCluster-2 with the approprate IP of the cluster
spring.cloud.stream.binders.kafka3.type: kstream
spring.cloud.stream.binders.kafka3.environment.spring.cloud.stream.kafka.streams.binder.brokers=${kafkaCluster-2} #Replace kafkaCluster-2 with the approprate IP of the cluster

spring.cloud.stream.function.definition=process;kstreamProcess

# From cluster 1 to cluster 2 with regular process function
spring.cloud.stream.bindings.process-in-0.destination=foo
spring.cloud.stream.bindings.process-in-0.binder=kafka1 # source from cluster 1
spring.cloud.stream.bindings.process-out-0.destination=bar
spring.cloud.stream.bindings.process-out-0.binder=kafka2 # send to cluster 2

# Kafka Streams processor on cluster 2
spring.cloud.stream.bindings.kstreamProcess-in-0.destination=bar
spring.cloud.stream.bindings.kstreamProcess-in-0.binder=kafka3
spring.cloud.stream.bindings.kstreamProcess-out-0.destination=foobar
spring.cloud.stream.bindings.kstreamProcess-out-0.binder=kafka3

Pay attention to the above configuration. We have two kinds of binders, but 3 binders all in all, first one is the regular Kafka binder based on cluster 1 (kafka1), then another Kafka binder based on cluster 2 (kafka2) and finally the kstream one (kafka3). The first processor in the application receives data from kafka1 and publishes to kafka2 where both binders are based on regular Kafka binder but differnt clusters. The second processor, which is a Kafka Streams processor consumes data from kafka3 which is the same cluster as kafka2, but a different binder type.

Since there are three different binder types available in the Kafka Streams family of binders - kstream, ktable and globalktable - if your application has multiple bindings based on any of these binders, that needs to be explicitly provided as the binder type.

For e.g if you have a processor as below,

@Bean
public Function<KStream<Long, Order>,
        Function<KTable<Long, Customer>,
                Function<GlobalKTable<Long, Product>, KStream<Long, EnrichedOrder>>>> enrichOrder() {

    ...
}

then, this has to be configured in a multi binder scenario as the following. Please note that this is only needed if you have a true multi-binder scenario where there are multiple processors dealing with multiple clusters within a single application. In that case, the binders need to be explicitly provided with the bindings to distinguish from other processor’s binder types and clusters.

spring.cloud.stream.binders.kafka1.type: kstream
spring.cloud.stream.binders.kafka1.environment.spring.cloud.stream.kafka.streams.binder.brokers=${kafkaCluster-2}
spring.cloud.stream.binders.kafka2.type: ktable
spring.cloud.stream.binders.kafka2.environment.spring.cloud.stream.kafka.streams.binder.brokers=${kafkaCluster-2}
spring.cloud.stream.binders.kafka3.type: globalktable
spring.cloud.stream.binders.kafka3.environment.spring.cloud.stream.kafka.streams.binder.brokers=${kafkaCluster-2}

spring.cloud.stream.bindings.enrichOrder-in-0.binder=kafka1  #kstream
spring.cloud.stream.bindings.enrichOrder-in-1.binder=kafka2  #ktablr
spring.cloud.stream.bindings.enrichOrder-in-2.binder=kafka3  #globalktable
spring.cloud.stream.bindings.enrichOrder-out-0.binder=kafka1 #kstream

# rest of the configuration is omitted.

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.

Kafka Streams topology visualization

Kafka Streams binder provides the following actuator endpoints for retrieving the topology description using which you can visualize the topology using external tools.

/actuator/kafkastreamstopology

/actuator/kafkastreamstopology/<application-id of the processor>

You need to include the actuator and web dependencies from Spring Boot to access these endpoints. Further, you also need to add kafkastreamstopology to management.endpoints.web.exposure.include property. By default, the kafkastreamstopology endpoint is disabled.

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.

Kafka Streams Binder Properties

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

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. All configuration that you can set from StreamsConfig can be set through this. When using this property, it is applicable against the entire application since this is a binder level property. If you have more than processors in the application, all of them will acquire these properties. In the case of properties like application.id, this will become problematic and therefore you have to carefully examine how the properties from StreamsConfig are mapped using this binder level configuration property.

functions.<function-bean-name>.applicationId: Applicable only for functional style processors. This can be used for setting application ID per function in the application. In the case of multiple functions, this is a handy way to set the application ID.
functions.<function-bean-name>.configuration: Applicable only for functional style processors. Map with a key/value pair containing properties pertaining to Apache Kafka Streams API. This is similar to the binder level configuration property describe above, but this level of configuration property is restricted only against the named function. When you have multiple processors and you want to restrict access to the configuration based on particular functions, you might want to use this. All StreamsConfig properties can be used here.
brokers: Broker URL

Default: localhost
zkNodes: Zookeeper URL

Default: localhost
deserializationExceptionHandler: Deserialization error handler type. This handler is applied at the binder level and thus applied against all input binding in the application. There is a way to control it in a more fine-grained way at the consumer binding level. 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 functions or StreamListener methods, then the application id should be set differently. See above where setting the application id is discussed in detail.

Default: application will generate a static application ID. See the application ID section for more details.
stateStoreRetry.maxAttempts: Max attempts for trying to connect to a state store.

Default: 1
stateStoreRetry.backoffPeriod: Backoff period when trying to connect to a state store on a retry.

Default: 1000 ms
consumerProperties: Arbitrary consumer properties at the binder level.
producerProperties: Arbitrary producer properties at the binder level.

Kafka Streams Producer Properties

The following properties are only available for Kafka Streams producers and must be prefixed with spring.cloud.stream.kafka.streams.bindings.<binding name>.producer. For convenience, if there are 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: See the above discussion on message de/serialization
valueSerde: value serde to use

Default: See the above discussion on message de/serialization
useNativeEncoding: flag to enable/disable native encoding

Default: true.

streamPartitionerBeanName: Custom outbound partitioner bean name to be used at the consumer. Applications can provide custom StreamPartitioner as a Spring bean and the name of this bean can be provided to the producer to use instead of the default one.

+ Default: See the discussion above on outbound partition support.

producedAs: Custom name for the sink component to which the processor is producing to.

Deafult: none (generated by Kafka Streams)

Kafka Streams Consumer Properties

The following properties are available for Kafka Streams consumers and must be prefixed with spring.cloud.stream.kafka.streams.bindings.<binding-name>.consumer. For convenience, if there are 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. This is only preferred for StreamListener based processors, for function based processors see other approaches outlined above.

Default: See above.
keySerde: key serde to use

Default: See the above discussion on message de/serialization
valueSerde: value serde to use

Default: See the above discussion on message de/serialization
materializedAs: state store to materialize when using incoming KTable types

Default: none.
useNativeDecoding: flag to enable/disable native decoding

Default: true.
dlqName: DLQ topic name.

Default: See above on the discussion of error handling and DLQ.
startOffset: Offset to start from if there is no committed offset to consume from. This is mostly used when the consumer is consuming from a topic for the first time. Kafka Streams uses earliest as the default strategy and the binder uses the same default. This can be overridden to latest using this property.

Default: earliest.

Note: Using resetOffsets on the consumer does not have any effect on Kafka Streams binder. Unlike the message channel based binder, Kafka Streams binder does not seek to beginning or end on demand.

deserializationExceptionHandler: Deserialization error handler type. This handler is applied per consumer binding as opposed to the binder level property described before. Possible values are - logAndContinue, logAndFail or sendToDlq

Default: logAndFail
timestampExtractorBeanName: Specific time stamp extractor bean name to be used at the consumer. Applications can provide TimestampExtractor as a Spring bean and the name of this bean can be provided to the consumer to use instead of the default one.

Default: See the discussion above on timestamp extractors.
eventTypes: Comma separated list of supported event types for this binding.

Default: none
eventTypeHeaderKey: Event type header key on each incoming records through this binding.

Default: event_type
consumedAs: Custom name for the source component from which the processor is consuming from.

Deafult: none (generated by Kafka Streams)

Special note on concurrency

In Kafka Streams, you can control of the number of threads a processor can create using the num.stream.threads property. This, you can do using the various configuration options described above under binder, functions, producer or consumer level. You can also use the concurrency property that core Spring Cloud Stream provides for this purpose. When using this, you need to use it on the consumer. When you have more than one input bindings either in a function or StreamListener, set this on the first input binding. For e.g. when setting spring.cloud.stream.bindings.process-in-0.consumer.concurrency, it will be translated as num.stream.threads by the binder. If you have multiple processors and one processor defines binding level concurrency, but not the others, those ones with no binding level concurrency will default back to the binder wide property specified through spring.cloud.stream.kafka.streams.binder.configuration.num.stream.threads. If this binder configuration is not available, then the application will use the default set by Kafka Streams.