How to create sliding windows

Question:

How do I create windowed calculations on time-series data with small advances in time?

Edit this page

Example use case:

You have time-series records and you want to create windowed aggregations with small increments in time. You could use hopping-windows, but with small time increments, hopping windows aren't the best solution with small increments in time. In this tutorial we'll cover how you can achieve efficient windowed aggregations with small advances in time

Code example:





Short Answer

To use SlidingWindows use the SlidingWindows.withTimeDifferenceAndGrace method inside a windowedBy call.



builder.stream(<INPUT TOPIC>, Consumed.with(<KEY SERDE>, <VALUE SERDE>))
                .groupByKey()
                .windowedBy(SlidingWindows.withTimeDifferenceAndGrace(Duration.ofSeconds(1), Duration.ofSeconds(1)))
                .<Aggergation Operation>....

The first parameter determinines the maximum time difference between records in the same window, and the second parameter sets the grace period for allowing out-of-order events into the window. For specifics on sliding windows you can read KIP-450.

You can get similar behavior in hopping windows by defining a short advance time. But this will result in poor performance because hopping windows will create many overlapping, possibly redundant windows. Performing aggregation operations over redundant windows costs CPU time, which can be expensive. Sliding windows only create windows containing distinct items, and perform calculations on these is more efficient. Additionally, sliding windows are inclusive on both the start and end time vs. hopping windows being inclusive only on the start time.

Try it

1
Initialize the project

To get started, make a new directory anywhere you’d like for this project:

mkdir sliding-windows && cd sliding-windows

2
Get Confluent Platform

Create a Dockerfile called Dockerfile-connect that builds a custom container for Kafka Connect bundled with the free and open source Kafka Connect Datagen connector, installed from Confluent Hub.

FROM confluentinc/cp-kafka-connect-base:6.1.0

ENV CONNECT_PLUGIN_PATH="/usr/share/java,/usr/share/confluent-hub-components"

RUN confluent-hub install --no-prompt confluentinc/kafka-connect-datagen:0.4.0

Next, create the following docker-compose.yml file to obtain Confluent Platform:

---
version: '2'

services:
  zookeeper:
    image: confluentinc/cp-zookeeper:6.1.0
    hostname: zookeeper
    container_name: zookeeper
    ports:
      - "2181:2181"
    environment:
      ZOOKEEPER_CLIENT_PORT: 2181
      ZOOKEEPER_TICK_TIME: 2000

  broker:
    image: confluentinc/cp-kafka:6.1.0
    hostname: broker
    container_name: broker
    depends_on:
      - zookeeper
    ports:
      - "29092:29092"
    environment:
      KAFKA_BROKER_ID: 1
      KAFKA_ZOOKEEPER_CONNECT: 'zookeeper:2181'
      KAFKA_LISTENER_SECURITY_PROTOCOL_MAP: PLAINTEXT:PLAINTEXT,PLAINTEXT_HOST:PLAINTEXT
      KAFKA_ADVERTISED_LISTENERS: PLAINTEXT://broker:9092,PLAINTEXT_HOST://localhost:29092
      KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR: 1
      KAFKA_GROUP_INITIAL_REBALANCE_DELAY_MS: 0

  schema-registry:
    image: confluentinc/cp-schema-registry:6.1.0
    hostname: schema-registry
    container_name: schema-registry
    depends_on:
      - broker
    ports:
      - "8081:8081"
    environment:
      SCHEMA_REGISTRY_HOST_NAME: schema-registry
      SCHEMA_REGISTRY_KAFKASTORE_BOOTSTRAP_SERVERS: 'broker:9092'

  connect:
    image: localimage/kafka-connect-datagen:latest
    build:
      context: .
      dockerfile: Dockerfile-connect
    container_name: connect
    depends_on:
      - broker
      - schema-registry
    ports:
      - 8083:8083
    volumes:
      - ./datagen-temperature-reading.avsc:/tmp/datagen-temperature-reading.avsc
    environment:
      CONNECT_BOOTSTRAP_SERVERS: "broker:9092"
      CONNECT_REST_ADVERTISED_HOST_NAME: "connect"
      CONNECT_REST_PORT: 8083
      CONNECT_GROUP_ID: kafka-connect
      CONNECT_CONFIG_STORAGE_TOPIC: _kafka-connect-configs
      CONNECT_OFFSET_STORAGE_TOPIC: _kafka-connect-offsets
      CONNECT_STATUS_STORAGE_TOPIC: _kafka-connect-status
      CONNECT_KEY_CONVERTER: org.apache.kafka.connect.storage.StringConverter
      CONNECT_VALUE_CONVERTER: io.confluent.connect.avro.AvroConverter
      CONNECT_VALUE_CONVERTER_SCHEMA_REGISTRY_URL: 'http://schema-registry:8081'
      CONNECT_LOG4J_ROOT_LOGLEVEL: "INFO"
      CONNECT_LOG4J_LOGGERS: "org.apache.kafka.connect.runtime.rest=WARN,org.reflections=ERROR"
      CONNECT_CONFIG_STORAGE_REPLICATION_FACTOR: "1"
      CONNECT_OFFSET_STORAGE_REPLICATION_FACTOR: "1"
      CONNECT_STATUS_STORAGE_REPLICATION_FACTOR: "1"

Now launch Confluent Platform by running the following command. Note the --build argument which automatically builds the Docker image for Kafka Connect and the bundled kafka-connect-datagen connector.

docker-compose up -d --build

3
Configure the project

Create the following Gradle build file, named build.gradle for the project:

buildscript {
    repositories {
        mavenCentral()
    }
    dependencies {
        classpath "com.commercehub.gradle.plugin:gradle-avro-plugin:0.22.0"
        classpath "com.github.jengelman.gradle.plugins:shadow:6.0.0"
    }
}

plugins {
    id "java"
    id "com.google.cloud.tools.jib" version "3.1.1"
    id "idea"
    id "eclipse"
}

sourceCompatibility = "1.8"
targetCompatibility = "1.8"
version = "0.0.1"

repositories {
    mavenCentral()

    maven {
        url "https://packages.confluent.io/maven"
    }
}

apply plugin: "com.commercehub.gradle.plugin.avro"
apply plugin: "com.github.johnrengelman.shadow"

dependencies {
    implementation "org.apache.avro:avro:1.10.2"
    implementation "org.slf4j:slf4j-simple:1.7.30"
    implementation "org.apache.kafka:kafka-streams:2.7.0"
    implementation('io.confluent:kafka-streams-avro-serde:6.1.1')
    testImplementation "org.apache.kafka:kafka-streams-test-utils:2.7.0"
    testImplementation "junit:junit:4.13.2"
    testImplementation 'org.hamcrest:hamcrest:2.2'
}

test {
    testLogging {
        outputs.upToDateWhen { false }
        showStandardStreams = true
        exceptionFormat = "full"
    }
}

jar {
  manifest {
    attributes(
      "Class-Path": configurations.compileClasspath.collect { it.getName() }.join(" "),
      "Main-Class": "io.confluent.developer.SlidingWindow"
    )
  }
}

shadowJar {
    archiveBaseName = "sliding-windows-standalone"
    archiveClassifier = ''
}

And be sure to run the following command to obtain the Gradle wrapper:

gradle wrapper

Next, create a directory for configuration data:

mkdir configuration

Then create a development file at configuration/dev.properties:

application.id=sliding-windows
bootstrap.servers=localhost:29092
schema.registry.url=http://localhost:8081

input.topic.name=temp-readings
input.topic.partitions=1
input.topic.replication.factor=1

output.topic.name=output-topic
output.topic.partitions=1
output.topic.replication.factor=1

4
Create a schema for the model object

Create a directory for the schemas that represent the events in the stream:

mkdir -p src/main/avro

Then create the following Avro schema file at src/main/avro/temperature_reading.avsc for our TemperatureReading object:

{
  "namespace": "io.confluent.developer.avro",
  "type": "record",
  "name": "TemperatureReading",
  "fields": [
    { "name": "temp", "type": "double" },
    { "name": "timestamp", "type": "long" } ,
    { "name": "device_id", "type": "string" }
  ]
}

You’ll also need to create another schema src/main/avro/temp_average.avsc for the TempAverage object that Kafka Streams will use for holding the data needed to perform the aggregation.

{
  "namespace": "io.confluent.developer.avro",
  "type": "record",
  "name": "TempAverage",
  "fields": [
    {
      "name": "total",
      "type": "double"
    },
    {
      "name": "num_readings",
      "type": "long"
    }
  ]
}

Because we will use an Avro schema in our Java code, we’ll need to compile it. The Gradle Avro plugin is a part of the build, so it will see your new Avro files, generate Java code for them, and compile those and all other Java sources. Run this command to get it all done:

./gradlew build

5
Create a timestamp extractor

First, create a directory for the Java files in this project:

mkdir -p src/main/java/io/confluent/developer

Before you create the Kafka Streams application you’ll need to create an instance of a TimestampExtractor. In Kafka Streams, timestamps drive the progress of records in the application. By default, Kafka Streams uses the timestamps contained in the ConsumerRecord. But you can configure your application to use timestamps embedded in the record payload itself. You do this by creating an class implementing the TimestampExtractor interface and provide the class name when configuring your Kafka Streams application like so:

 props.put(StreamsConfig.DEFAULT_TIMESTAMP_EXTRACTOR_CLASS_CONFIG, TemperatureReadingTimestampExtractor.class.getName());

We’re going to create a custom TimestampExtractor so the Kafka Streams application uses the timestamps embedded in our generated IoT sensor readings.

Create the following file at src/main/java/io/confluent/developer/TemperatureReadingTimestampExtractor.java

package io.confluent.developer;

import io.confluent.developer.avro.TemperatureReading;
import org.apache.kafka.clients.consumer.ConsumerRecord;
import org.apache.kafka.streams.processor.TimestampExtractor;

public class TemperatureReadingTimestampExtractor implements TimestampExtractor {
    @Override
    public long extract(ConsumerRecord<Object, Object> record, long previousTimestamp) {
        return ((TemperatureReading)record.value()).getTimestamp();
    }
}

You’ll take care of the configuration when you create the Kafka Streams topology in the next step.

6
Create the Kafka Streams topology

Before you create the Kafka Streams application let’s go over the key points. In this tutorial you’ll learn about using SlidingWindows for aggregation operations. Incoming record timestamps drive the window behavior, and the time difference between records plus a grace period for out-of-order records (both are user defined) drives the size of the window.

   builder.stream(tempReadingTopic, Consumed.with(Serdes.String(), temReadingSerde))
                .groupByKey() (1)
                .windowedBy(SlidingWindows.withTimeDifferenceAndGrace(Duration.ofSeconds(1), Duration.ofSeconds(1))) (2)
                .aggregate(TempAverage::new, (k, tr, ave) -> { (3)
                    ave.setNumReadings(ave.getNumReadings() +1);
                    ave.setTotal(ave.getTotal() + tr.getTemp());
                    return ave;
                }, Materialized.with(Serdes.String(), tempAveSerde))
1 Grouping by key, a prerequisite for aggregation operations
2 Specifying a sliding window for the windowing operation
3 The aggregation operation which sums the temperature and keeps a total count of the readings

As the window "slides" over the record stream, a new window is created each time a record enters or exits the window. Records that arrive after grace period are dropped. A window closes when stream-time exceeds the window-end + grace-period

You can get similar behavior in hopping windows by defining a short advance time. But this will result in poor performance because hopping windows will create many overlapping, possibly redundant windows. Performing aggregation operations over redundant windows costs CPU time, which can be expensive. Sliding windows only create windows containing distinct items, and perform calculations on these is more efficient. Additionally, sliding windows are inclusive on both the start and end time vs. hopping windows being inclusive only on the start time.

The concept of stream-time is the largest timestamp processed by Kafka Streams at a given point in time. A grace-period is the amount of time you’re allowing out-of-order records to be included in a window.

That wraps up our discussion for the finer points of the code for this tutorial. Now create the following file at src/main/java/io/confluent/developer/SlidingWindow.java

package io.confluent.developer;

import io.confluent.developer.avro.TempAverage;
import io.confluent.developer.avro.TemperatureReading;
import io.confluent.kafka.serializers.AbstractKafkaSchemaSerDeConfig;
import io.confluent.kafka.streams.serdes.avro.SpecificAvroSerde;
import org.apache.avro.specific.SpecificRecord;
import org.apache.kafka.clients.admin.AdminClient;
import org.apache.kafka.clients.admin.NewTopic;
import org.apache.kafka.common.serialization.Serdes;
import org.apache.kafka.streams.KafkaStreams;
import org.apache.kafka.streams.KeyValue;
import org.apache.kafka.streams.StreamsBuilder;
import org.apache.kafka.streams.StreamsConfig;
import org.apache.kafka.streams.Topology;
import org.apache.kafka.streams.kstream.Consumed;
import org.apache.kafka.streams.kstream.Materialized;
import org.apache.kafka.streams.kstream.Produced;
import org.apache.kafka.streams.kstream.SlidingWindows;
import org.apache.kafka.streams.kstream.Windowed;

import java.io.FileInputStream;
import java.io.IOException;
import java.time.Duration;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Properties;
import java.util.concurrent.CountDownLatch;

import static io.confluent.kafka.serializers.AbstractKafkaSchemaSerDeConfig.SCHEMA_REGISTRY_URL_CONFIG;

public class SlidingWindow {

    public Properties buildStreamsProperties(Properties envProps) {
        Properties props = new Properties();

        props.put(StreamsConfig.APPLICATION_ID_CONFIG, envProps.getProperty("application.id"));
        props.put(StreamsConfig.BOOTSTRAP_SERVERS_CONFIG, envProps.getProperty("bootstrap.servers"));
        props.put(StreamsConfig.DEFAULT_KEY_SERDE_CLASS_CONFIG, Serdes.String().getClass());
        props.put(StreamsConfig.DEFAULT_VALUE_SERDE_CLASS_CONFIG, SpecificAvroSerde.class);
        props.put(SCHEMA_REGISTRY_URL_CONFIG, envProps.getProperty("schema.registry.url"));
        props.put(StreamsConfig.DEFAULT_TIMESTAMP_EXTRACTOR_CLASS_CONFIG, TemperatureReadingTimestampExtractor.class.getName());
        // Setting this to a low value on purpose to flush the cache quickly during the demo
        // Normally you'll want to keep this setting at the default value of 30 seconds
        props.put(StreamsConfig.COMMIT_INTERVAL_MS_CONFIG, 1000);

        return props;
    }

    public Topology buildTopology(Properties envProps) {
        final StreamsBuilder builder = new StreamsBuilder();
        final String tempReadingTopic = envProps.getProperty("input.topic.name");
        final String aveTempOutputTopic = envProps.getProperty("output.topic.name");
        final SpecificAvroSerde<TemperatureReading> temReadingSerde = getSpecificAvroSerde(envProps);
        final SpecificAvroSerde<TempAverage> tempAveSerde = getSpecificAvroSerde(envProps);

        builder.stream(tempReadingTopic, Consumed.with(Serdes.String(), temReadingSerde))
                .groupByKey()
                .windowedBy(SlidingWindows.withTimeDifferenceAndGrace(Duration.ofMillis(500), Duration.ofMillis(100)))
                .aggregate(TempAverage::new, (k, tr, ave) -> {
                    ave.setNumReadings(ave.getNumReadings() +1);
                    ave.setTotal(ave.getTotal() + tr.getTemp());
                    return ave;
                }, Materialized.with(Serdes.String(), tempAveSerde))
                .toStream()
                .map((Windowed<String> key, TempAverage tempAverage) -> {
                    double aveNoFormat = tempAverage.getTotal()/(double)tempAverage.getNumReadings();
                    double formattedAve = Double.parseDouble(String.format("%.2f", aveNoFormat));
                    return new KeyValue<>(key.key(),formattedAve) ;
                })
                .to(aveTempOutputTopic, Produced.with(Serdes.String(), Serdes.Double()));

        return builder.build();
    }


    static <T extends SpecificRecord> SpecificAvroSerde<T> getSpecificAvroSerde(final Properties envProps) {
        final SpecificAvroSerde<T> specificAvroSerde = new SpecificAvroSerde<>();

        final HashMap<String, String> serdeConfig = new HashMap<>();
        serdeConfig.put(AbstractKafkaSchemaSerDeConfig.SCHEMA_REGISTRY_URL_CONFIG,
                envProps.getProperty("schema.registry.url"));

        specificAvroSerde.configure(serdeConfig, false);
        return specificAvroSerde;
    }
    public void createTopics(Properties envProps) {
        Map<String, Object> config = new HashMap<>();
        config.put("bootstrap.servers", envProps.getProperty("bootstrap.servers"));
        try (AdminClient client = AdminClient.create(config)) {
            List<NewTopic> topics = new ArrayList<>();

            NewTopic ratings = new NewTopic(envProps.getProperty("input.topic.name"),
                    Integer.parseInt(envProps.getProperty("input.topic.partitions")),
                    Short.parseShort(envProps.getProperty("input.topic.replication.factor")));
            topics.add(ratings);

            NewTopic counts = new NewTopic(envProps.getProperty("output.topic.name"),
                    Integer.parseInt(envProps.getProperty("output.topic.partitions")),
                    Short.parseShort(envProps.getProperty("output.topic.replication.factor")));

            topics.add(counts);
            client.createTopics(topics);
        }
    }

    public Properties loadEnvProperties(String fileName) throws IOException {
        Properties envProps = new Properties();
        FileInputStream input = new FileInputStream(fileName);
        envProps.load(input);
        input.close();

        return envProps;
    }

    public static void main(String[] args) throws Exception {

        if (args.length < 1) {
            throw new IllegalArgumentException("This program takes one argument: the path to an environment configuration file.");
        }

        SlidingWindow tw = new SlidingWindow();
        Properties envProps = tw.loadEnvProperties(args[0]);
        Properties streamProps = tw.buildStreamsProperties(envProps);
        Topology topology = tw.buildTopology(envProps);

        tw.createTopics(envProps);

        final KafkaStreams streams = new KafkaStreams(topology, streamProps);
        final CountDownLatch latch = new CountDownLatch(1);

        // Attach shutdown handler to catch Control-C.
        Runtime.getRuntime().addShutdownHook(new Thread("streams-shutdown-hook") {
            @Override
            public void run() {
                streams.close();
                latch.countDown();
            }
        });

        try {
            streams.cleanUp();
            streams.start();
            latch.await();
        } catch (Throwable e) {
            System.exit(1);
        }
        System.exit(0);
    }
}

7
Start data generation for the Kafka Streams application

Before you start your Kafka Streams application, we need to provide data for it. Fortunately this is as simple as using a HTTP PUT request, as you’re going to use the DatagenConnector.

Now create the following Avro schema file datagen-temperature-reading.avsc in the current working directory (sliding-windows) for the tutorial:

{
  "namespace": "io.confluent.developer.avro",
  "type": "record",
  "name": "TemperatureReading",
  "fields": [
    { "name": "temp", "type": {
        "type": "double",
        "arg.properties": {
          "range": { "min": 70.00, "max": 99.91}
        }
      }
    },
    { "name": "timestamp", "type": {
      "type": "long",
      "format_as_time" : "unix_long",
      "arg.properties": {
        "iteration": { "start": 1, "step": 100}
      }
    }},
    { "name": "device_id", "type": {
        "type": "string",
        "arg.properties": {
          "options": ["device-1", "device-2"]
        }
      }
    }
  ]
}

This schema file is pretty much idential to the one you created earlier.

The only difference is this schema contains instructions for data generation. The kafka-connect-datagen connector uses the Avro Random Generator to generate data.

Open an new terminal window and run this command to start the data generator:

curl -i -X PUT http://localhost:8083/connectors/datagen_local_01/config \
     -H "Content-Type: application/json" \
     -d '{
            "connector.class": "io.confluent.kafka.connect.datagen.DatagenConnector",
            "key.converter": "org.apache.kafka.connect.storage.StringConverter",
            "kafka.topic": "temp-readings",
            "schema.filename": "/tmp/datagen-temperature-reading.avsc",
            "schema.keyfield": "device_id",
            "max.interval": 300,
            "iterations": 10000000,
            "tasks.max": "1"
        }'

You should see something like this on the console indicating the datagen connector sucessfuly started

HTTP/1.1 201 Created
Date: Tue, 26 Jan 2021 21:40:33 GMT
Location: http://localhost:8083/connectors/datagen_local_01
Content-Type: application/json
Content-Length: 413
Server: Jetty(9.4.24.v20191120)

{"name":"datagen_local_01","config":{"connector.class":"io.confluent.kafka.connect.datagen.DatagenConnector","key.converter":"org.apache.kafka.connect.storage.StringConverter","kafka.topic":"temp-readings","schema.filename":"/schemas/datagen-temperature-reading.avsc","schema.keyfield":"device_id","max.interval":"300","iterations":"10000000","tasks.max":"1","name":"datagen_local_01"},"tasks":[],"type":"source"}

8
Compile and run the Kafka Streams program

Now that we have data generation working, let’s build your application by running:

./gradlew shadowJar

Now that you have an uberjar for the Kafka Streams application, you can launch it locally. When you run the following, the prompt won’t return, because the application will run until you exit it. There is always another message to process, so streaming applications don’t exit until you force them.

java -jar build/libs/sliding-windows-standalone-0.0.1.jar configuration/dev.properties

9
Consume data from the output topic

Now that your Kafka Streams application is running, open a new terminal window, change directories (cd) into the sliding-windows directory and start a console-consumer to confirm the output:

docker-compose exec broker kafka-console-consumer \
 --bootstrap-server broker:9092 \
 --topic output-topic \
 --property print.key=true \
 --value-deserializer "org.apache.kafka.common.serialization.DoubleDeserializer" \
 --property key.separator=" : "  \
 --from-beginning \
 --max-messages 10

Your results should look someting like this:


device-1 : 79.77
device-1 : 80.63
device-2 : 89.87
device-1 : 79.76
device-1 : 80.06
device-1 : 82.06
device-2 : 73.63
device-2 : 81.75
device-1 : 86.32
device-1 : 82.82

Go ahead and shut down your streams application now with a CNTR+C command.

Test it

1
Create a test configuration file

First, create a test file at configuration/test.properties:

application.id=sliding-windows-test
bootstrap.servers=localhost:29092
schema.registry.url=mock://localhost:8081

input.topic.name=temp-readings
input.topic.partitions=1
input.topic.replication.factor=1

output.topic.name=output-topic
output.topic.partitions=1
output.topic.replication.factor=1

2
Write a test

Create a directory for the tests to live in:

mkdir -p src/test/java/io/confluent/developer

Testing a Kafka streams application requires a bit of test harness code, but happily the org.apache.kafka.streams.TopologyTestDriver class makes this much more pleasant that it would otherwise be.

There is only one method in SlidingWindowTest annotated with @Test, and that is slidingWindowTest(). This method actually runs our Streams topology using the TopologyTestDriver and some mocked data that is set up inside the test method.

This test is straightforward, but there is one section we should look into a little more

final String key = "device-1";
final List<TemperatureReading> temperatureReadings = new ArrayList<>();
Instant instant = Instant.now().truncatedTo(ChronoUnit.MINUTES);

temperatureReadings.add(TemperatureReading.newBuilder().setDeviceId(key).setTemp(80.0).setTimestamp(instant.getEpochSecond()).build());    (1)
temperatureReadings.add(TemperatureReading.newBuilder().setDeviceId(key).setTemp(90.0).setTimestamp(instant.plusMillis(200).getEpochSecond()).build());
temperatureReadings.add(TemperatureReading.newBuilder().setDeviceId(key).setTemp(95.0).setTimestamp(instant.plusMillis(400).getEpochSecond()).build());
temperatureReadings.add(TemperatureReading.newBuilder().setDeviceId(key).setTemp(100.0).setTimestamp(instant.plusMillis(500).getEpochSecond()).build());

List<KeyValue<String, TemperatureReading>> keyValues = temperatureReadings.stream().map(o -> KeyValue.pair(o.getDeviceId(),o)).collect(Collectors.toList()); (2)
inputTopic.pipeKeyValueList(keyValues);
1 Creating the input records for the test
2 Mapping the list of test records to KeyValue pairs

The TestInputTopic provides useful methods when testing your topology. Here you’re using the pipeKeyValueList to provide the records to the steams application. Here you’re not specifying any timestamp activity as the streams application pulls the timestamps embedded in the TemperatureReading objects you created above.

Now create the following file at src/test/java/io/confluent/developer/SlidingWindowTest.java.

package io.confluent.developer;


import io.confluent.common.utils.TestUtils;
import io.confluent.developer.avro.TemperatureReading;
import io.confluent.kafka.streams.serdes.avro.SpecificAvroSerde;
import org.apache.kafka.common.serialization.Deserializer;
import org.apache.kafka.common.serialization.Serdes;
import org.apache.kafka.common.serialization.Serializer;
import org.apache.kafka.streams.KeyValue;
import org.apache.kafka.streams.StreamsConfig;
import org.apache.kafka.streams.TestInputTopic;
import org.apache.kafka.streams.TestOutputTopic;
import org.apache.kafka.streams.Topology;
import org.apache.kafka.streams.TopologyTestDriver;
import org.junit.After;
import org.junit.Test;
import static org.junit.Assert.*;

import java.io.IOException;
import java.nio.file.Files;
import java.time.Instant;
import java.time.temporal.ChronoUnit;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.List;
import java.util.Properties;
import java.util.stream.Collectors;


public class SlidingWindowTest {

    private final static String TEST_CONFIG_FILE = "configuration/test.properties";

    @Test
    public void slidingWindowTest() throws IOException {
        final SlidingWindow instance = new SlidingWindow();
        final Properties envProps = instance.loadEnvProperties(TEST_CONFIG_FILE);

        final Properties streamProps = instance.buildStreamsProperties(envProps);
        final String temperatureReadingsInputTopic = envProps.getProperty("input.topic.name");
        final String outputTopicName = envProps.getProperty("output.topic.name");

        final Topology topology = instance.buildTopology(envProps);
        try (final TopologyTestDriver testDriver = new TopologyTestDriver(topology, streamProps)) {

            final SpecificAvroSerde<TemperatureReading> exampleAvroSerde = SlidingWindow.getSpecificAvroSerde(envProps);

            final Serializer<String> keySerializer = Serdes.String().serializer();
            final Serializer<TemperatureReading> exampleSerializer = exampleAvroSerde.serializer();
            final Deserializer<Double> valueDeserializer = Serdes.Double().deserializer();
            final Deserializer<String> keyDeserializer = Serdes.String().deserializer();

            final TestInputTopic<String, TemperatureReading>  inputTopic = testDriver.createInputTopic(temperatureReadingsInputTopic,
                                                                                              keySerializer,
                                                                                              exampleSerializer);

            final TestOutputTopic<String, Double> outputTopic = testDriver.createOutputTopic(outputTopicName, keyDeserializer, valueDeserializer);
            final String key = "device-1";
            final List<TemperatureReading> temperatureReadings = new ArrayList<>();
            Instant instant = Instant.now().truncatedTo(ChronoUnit.MINUTES);
            temperatureReadings.add(TemperatureReading.newBuilder().setDeviceId(key).setTemp(80.0).setTimestamp(instant.getEpochSecond()).build());
            temperatureReadings.add(TemperatureReading.newBuilder().setDeviceId(key).setTemp(90.0).setTimestamp(instant.plusMillis(200).getEpochSecond()).build());
            temperatureReadings.add(TemperatureReading.newBuilder().setDeviceId(key).setTemp(95.0).setTimestamp(instant.plusMillis(400).getEpochSecond()).build());
            temperatureReadings.add(TemperatureReading.newBuilder().setDeviceId(key).setTemp(100.0).setTimestamp(instant.plusMillis(500).getEpochSecond()).build());

            List<KeyValue<String, TemperatureReading>> keyValues = temperatureReadings.stream().map(o -> KeyValue.pair(o.getDeviceId(),o)).collect(Collectors.toList());
            inputTopic.pipeKeyValueList(keyValues);
            final List<KeyValue<String, Double>> expectedValues = Arrays.asList(KeyValue.pair(key, 80.0), KeyValue.pair(key, 85.0), KeyValue.pair(key, 88.33), KeyValue.pair(key, 91.25));

            final List<KeyValue<String, Double>> actualResults = outputTopic.readKeyValuesToList();
            assertEquals(expectedValues, actualResults);
        }
    }
}

3
Invoke the tests

Now run the test, which is as simple as:

./gradlew test

Take it to production

1
Create a production configuration file

First, create a new configuration file at configuration/prod.properties with the following content. Be sure to fill in the addresses of your production hosts and change any other parameters that make sense for your setup.

application.id=sliding-windows
bootstrap.servers=<<FILL ME IN>>
schema.registry.url=<<FILL ME IN>>

input.topic.name=temp-readings
input.topic.partitions=<<FILL ME IN>>
input.topic.replication.factor=<<FILL ME IN>>

output.topic.name=output-topic
output.topic.partitions=<<FILL ME IN>>
output.topic.replication.factor=<<FILL ME IN>>

2
Build a Docker image

In your terminal, execute the following to invoke the Jib plugin to build an image:

gradle jibDockerBuild --image=io.confluent.developer/sliding-windows-join:0.0.1

3
Launch the container

Finally, launch the container using your preferred container orchestration service. If you want to run it locally, you can execute the following:

docker run -v $PWD/configuration/prod.properties:/config.properties io.confluent.developer/sliding-windows-join:0.0.1 config.properties

Deploy on Confluent Cloud

1
Run your app to Confluent Cloud

Instead of running a local Kafka cluster, you may use Confluent Cloud, a fully-managed Apache Kafka service.

First, create your Kafka cluster in Confluent Cloud. Use the promo code CC100KTS to receive an additional $100 free usage (details).

Next, from the Confluent Cloud UI, click on Tools & client config to get the cluster-specific configurations, e.g. Kafka cluster bootstrap servers and credentials, Confluent Cloud Schema Registry and credentials, etc., and set the appropriate parameters in your client application. In the case of this tutorial, add the following properties to the client application’s input properties file, substituting all curly braces with your Confluent Cloud values.

# Required connection configs for Kafka producer, consumer, and admin
bootstrap.servers={{ BROKER_ENDPOINT }}
security.protocol=SASL_SSL
sasl.jaas.config=org.apache.kafka.common.security.plain.PlainLoginModule required username='{{ CLUSTER_API_KEY }}' password='{{ CLUSTER_API_SECRET }}';
sasl.mechanism=PLAIN
# Required for correctness in Apache Kafka clients prior to 2.6
client.dns.lookup=use_all_dns_ips

# Best practice for Kafka producer to prevent data loss
acks=all

# Required connection configs for Confluent Cloud Schema Registry
schema.registry.url=https://{{ SR_ENDPOINT }}
basic.auth.credentials.source=USER_INFO
schema.registry.basic.auth.user.info={{ SR_API_KEY }}:{{ SR_API_SECRET }}

Now you’re all set to run your streaming application locally, backed by a Kafka cluster fully managed by Confluent Cloud.