From: Kai Moritz Date: Sun, 24 Jul 2022 13:35:14 +0000 (+0200) Subject: Merge der Refaktorisierung des EndlessConsumer (Branch 'stored-state') X-Git-Tag: endless-stream-consumer-DEPRECATED^2^2^2~1^2~12 X-Git-Url: http://juplo.de/gitweb/?a=commitdiff_plain;h=e53004f3133b737699e995a3b18fff28203a0e8c;p=demos%2Fkafka%2Ftraining Merge der Refaktorisierung des EndlessConsumer (Branch 'stored-state') * Die `commtSync()`-Aufrufe machen beim Speichern der Offsets außerhalb von Kafka keinen Sinn mehr. * Der Testfall musste an die extern gespeicherten Offsets angepasst werden: Die gesehenen Offsets müssen aus der MongoDB gelesen werden, anstatt über einen separaten Consumer aus Kafka. * Der mit dem Merge hinzugefügte Test schlägt fehl, da er einen Fehler aufdeckt (NPE bei einer Log-Ausgabe zur Offset-Verarbeitung). --- e53004f3133b737699e995a3b18fff28203a0e8c diff --cc src/main/java/de/juplo/kafka/ApplicationConfiguration.java index 0000000,1ba9d5b..54e9b89 mode 000000,100644..100644 --- a/src/main/java/de/juplo/kafka/ApplicationConfiguration.java +++ b/src/main/java/de/juplo/kafka/ApplicationConfiguration.java @@@ -1,0 -1,69 +1,70 @@@ + package de.juplo.kafka; + + import org.apache.kafka.clients.consumer.ConsumerRecord; + import org.apache.kafka.clients.consumer.KafkaConsumer; + import org.apache.kafka.common.serialization.LongDeserializer; + import org.apache.kafka.common.serialization.StringDeserializer; + import org.springframework.boot.context.properties.EnableConfigurationProperties; + import org.springframework.context.annotation.Bean; + import org.springframework.context.annotation.Configuration; + + import java.util.Properties; + import java.util.concurrent.ExecutorService; + import java.util.concurrent.Executors; + import java.util.function.Consumer; + + + @Configuration + @EnableConfigurationProperties(ApplicationProperties.class) + public class ApplicationConfiguration + { + @Bean + public Consumer> consumer() + { + return (record) -> + { + // Handle record + }; + } + + @Bean + public EndlessConsumer endlessConsumer( + KafkaConsumer kafkaConsumer, + ExecutorService executor, + Consumer> handler, + PartitionStatisticsRepository repository, + ApplicationProperties properties) + { + return + new EndlessConsumer<>( + executor, + repository, + properties.getClientId(), + properties.getTopic(), + kafkaConsumer, + handler); + } + + @Bean + public ExecutorService executor() + { + return Executors.newSingleThreadExecutor(); + } + + @Bean(destroyMethod = "close") + public KafkaConsumer kafkaConsumer(ApplicationProperties properties) + { + Properties props = new Properties(); + + props.put("bootstrap.servers", properties.getBootstrapServer()); + props.put("group.id", properties.getGroupId()); + props.put("client.id", properties.getClientId()); ++ props.put("enable.auto.commit", false); + props.put("auto.offset.reset", properties.getAutoOffsetReset()); + props.put("metadata.max.age.ms", "1000"); + props.put("key.deserializer", StringDeserializer.class.getName()); + props.put("value.deserializer", LongDeserializer.class.getName()); + + return new KafkaConsumer<>(props); + } + } diff --cc src/main/java/de/juplo/kafka/EndlessConsumer.java index c79b1e0,a21dd86..2a3445c --- a/src/main/java/de/juplo/kafka/EndlessConsumer.java +++ b/src/main/java/de/juplo/kafka/EndlessConsumer.java @@@ -35,30 -33,59 +33,58 @@@ public class EndlessConsumer impl private boolean running = false; private Exception exception; private long consumed = 0; - private KafkaConsumer consumer = null; + private final Map> seen = new HashMap<>(); + private final Map offsets = new HashMap<>(); - private final Map> seen = new HashMap<>(); + @Override + public void onPartitionsRevoked(Collection partitions) + { + partitions.forEach(tp -> + { + Integer partition = tp.partition(); + Long newOffset = consumer.position(tp); + Long oldOffset = offsets.remove(partition); + log.info( + "{} - removing partition: {}, consumed {} records (offset {} -> {})", + id, + partition, + newOffset - oldOffset, + oldOffset, + newOffset); + Map removed = seen.remove(partition); + for (String key : removed.keySet()) + { + log.info( + "{} - Seen {} messages for partition={}|key={}", + id, + removed.get(key), + partition, + key); + } - repository.save(new StatisticsDocument(partition, removed)); ++ repository.save(new StatisticsDocument(partition, removed, consumer.position(tp))); + }); + } - public EndlessConsumer( - ExecutorService executor, - PartitionStatisticsRepository repository, - String bootstrapServer, - String groupId, - String clientId, - String topic, - String autoOffsetReset) + @Override + public void onPartitionsAssigned(Collection partitions) { - this.executor = executor; - this.repository = repository; - this.bootstrapServer = bootstrapServer; - this.groupId = groupId; - this.id = clientId; - this.topic = topic; - this.autoOffsetReset = autoOffsetReset; + partitions.forEach(tp -> + { + Integer partition = tp.partition(); + Long offset = consumer.position(tp); + log.info("{} - adding partition: {}, offset={}", id, partition, offset); - offsets.put(partition, offset); - seen.put( - partition, ++ StatisticsDocument document = + repository - .findById(Integer.toString(tp.partition())) - .map(document -> document.statistics) - .orElse(new HashMap<>())); ++ .findById(Integer.toString(partition)) ++ .orElse(new StatisticsDocument(partition)); ++ consumer.seek(tp, document.offset); ++ seen.put(partition, document.statistics); + }); } + @Override public void run() { @@@ -156,9 -132,24 +137,22 @@@ } catch(WakeupException e) { - log.info("{} - RIIING!", id); + log.info("{} - RIIING! Request to stop consumption - commiting current offsets!", id); - consumer.commitSync(); shutdown(); } + catch(RecordDeserializationException e) + { + TopicPartition tp = e.topicPartition(); + long offset = e.offset(); + log.error( + "{} - Could not deserialize message on topic {} with offset={}: {}", + id, + tp, + offset, + e.getCause().toString()); + - consumer.commitSync(); + shutdown(e); + } catch(Exception e) { log.error("{} - Unexpected error: {}", id, e.toString(), e); diff --cc src/main/java/de/juplo/kafka/StatisticsDocument.java index 96ebfb1,2416253..28264ec --- a/src/main/java/de/juplo/kafka/StatisticsDocument.java +++ b/src/main/java/de/juplo/kafka/StatisticsDocument.java @@@ -14,20 -14,13 +14,20 @@@ public class StatisticsDocumen { @Id public String id; + public long offset; - public Map statistics; + public Map statistics; public StatisticsDocument() { } - public StatisticsDocument(Integer partition, Map statistics) + public StatisticsDocument(Integer partition) + { + this.id = Integer.toString(partition); + this.statistics = new HashMap<>(); + } + - public StatisticsDocument(Integer partition, Map statistics, long offset) ++ public StatisticsDocument(Integer partition, Map statistics, long offset) { this.id = Integer.toString(partition); this.statistics = statistics; diff --cc src/test/java/de/juplo/kafka/ApplicationTests.java index 0000000,caa25c5..4b7ef36 mode 000000,100644..100644 --- a/src/test/java/de/juplo/kafka/ApplicationTests.java +++ b/src/test/java/de/juplo/kafka/ApplicationTests.java @@@ -1,0 -1,323 +1,313 @@@ + package de.juplo.kafka; + + import lombok.extern.slf4j.Slf4j; + import org.apache.kafka.clients.consumer.ConsumerRecord; + import org.apache.kafka.clients.consumer.KafkaConsumer; + import org.apache.kafka.clients.producer.KafkaProducer; + import org.apache.kafka.clients.producer.ProducerRecord; + import org.apache.kafka.common.TopicPartition; + import org.apache.kafka.common.errors.RecordDeserializationException; + import org.apache.kafka.common.serialization.*; + import org.apache.kafka.common.utils.Bytes; + import org.junit.jupiter.api.*; + import org.springframework.beans.factory.annotation.Autowired; + import org.springframework.boot.autoconfigure.EnableAutoConfiguration; + import org.springframework.boot.test.autoconfigure.data.mongo.AutoConfigureDataMongo; + import org.springframework.boot.test.context.ConfigDataApplicationContextInitializer; + import org.springframework.boot.test.context.TestConfiguration; + import org.springframework.context.annotation.Bean; + import org.springframework.context.annotation.Import; + import org.springframework.kafka.test.context.EmbeddedKafka; + import org.springframework.test.context.TestPropertySource; + import org.springframework.test.context.junit.jupiter.SpringJUnitConfig; + + import java.time.Duration; + import java.util.*; + import java.util.concurrent.ExecutionException; + import java.util.concurrent.ExecutorService; + import java.util.function.BiConsumer; + import java.util.function.Consumer; + import java.util.function.Function; + import java.util.stream.Collectors; + import java.util.stream.IntStream; + + import static de.juplo.kafka.ApplicationTests.PARTITIONS; + import static de.juplo.kafka.ApplicationTests.TOPIC; + import static org.assertj.core.api.Assertions.*; + import static org.awaitility.Awaitility.*; + + + @SpringJUnitConfig(initializers = ConfigDataApplicationContextInitializer.class) + @TestMethodOrder(MethodOrderer.OrderAnnotation.class) + @TestPropertySource( + properties = { + "consumer.bootstrap-server=${spring.embedded.kafka.brokers}", + "consumer.topic=" + TOPIC, + "spring.mongodb.embedded.version=4.4.13" }) + @EmbeddedKafka(topics = TOPIC, partitions = PARTITIONS) + @EnableAutoConfiguration + @AutoConfigureDataMongo + @Slf4j + class ApplicationTests + { + public static final String TOPIC = "FOO"; + public static final int PARTITIONS = 10; + + + StringSerializer stringSerializer = new StringSerializer(); + + @Autowired + Serializer valueSerializer; + @Autowired + KafkaProducer kafkaProducer; + @Autowired + KafkaConsumer kafkaConsumer; + @Autowired - KafkaConsumer offsetConsumer; ++ PartitionStatisticsRepository partitionStatisticsRepository; + @Autowired + ApplicationProperties properties; + @Autowired + ExecutorService executor; + @Autowired + PartitionStatisticsRepository repository; + + Consumer> testHandler; + EndlessConsumer endlessConsumer; + Map oldOffsets; + Map newOffsets; + Set> receivedRecords; + + + /** Tests methods */ + + @Test + @Order(1) // << The poistion pill is not skipped. Hence, this test must run first + void commitsCurrentOffsetsOnSuccess() throws ExecutionException, InterruptedException + { + send100Messages(i -> new Bytes(valueSerializer.serialize(TOPIC, i))); + + await("100 records received") + .atMost(Duration.ofSeconds(30)) + .until(() -> receivedRecords.size() >= 100); + + await("Offsets committed") + .atMost(Duration.ofSeconds(10)) + .untilAsserted(() -> + { + checkSeenOffsetsForProgress(); + compareToCommitedOffsets(newOffsets); + }); + + assertThatExceptionOfType(IllegalStateException.class) + .isThrownBy(() -> endlessConsumer.exitStatus()) + .describedAs("Consumer should still be running"); + } + + @Test + @Order(2) + void commitsOffsetOfErrorForReprocessingOnError() + { + send100Messages(counter -> + counter == 77 + ? new Bytes(stringSerializer.serialize(TOPIC, "BOOM!")) + : new Bytes(valueSerializer.serialize(TOPIC, counter))); + + await("Consumer failed") + .atMost(Duration.ofSeconds(30)) + .until(() -> !endlessConsumer.running()); + + checkSeenOffsetsForProgress(); + compareToCommitedOffsets(newOffsets); + + endlessConsumer.start(); + await("Consumer failed") + .atMost(Duration.ofSeconds(30)) + .until(() -> !endlessConsumer.running()); + + checkSeenOffsetsForProgress(); + compareToCommitedOffsets(newOffsets); + assertThat(receivedRecords.size()) + .describedAs("Received not all sent events") + .isLessThan(100); + + assertThatNoException() + .describedAs("Consumer should not be running") + .isThrownBy(() -> endlessConsumer.exitStatus()); + assertThat(endlessConsumer.exitStatus()) + .describedAs("Consumer should have exited abnormally") + .containsInstanceOf(RecordDeserializationException.class); + } + + + /** Helper methods for the verification of expectations */ + + void compareToCommitedOffsets(Map offsetsToCheck) + { + doForCurrentOffsets((tp, offset) -> + { + Long expected = offsetsToCheck.get(tp) + 1; + log.debug("Checking, if the offset for {} is {}", tp, expected); + assertThat(offset) + .describedAs("Committed offset corresponds to the offset of the consumer") + .isEqualTo(expected); + }); + } + + void checkSeenOffsetsForProgress() + { + // Be sure, that some messages were consumed...! + Set withProgress = new HashSet<>(); + partitions().forEach(tp -> + { + Long oldOffset = oldOffsets.get(tp); + Long newOffset = newOffsets.get(tp); + if (!oldOffset.equals(newOffset)) + { + log.debug("Progress for {}: {} -> {}", tp, oldOffset, newOffset); + withProgress.add(tp); + } + }); + assertThat(withProgress) + .describedAs("Some offsets must have changed, compared to the old offset-positions") + .isNotEmpty(); + } + + + /** Helper methods for setting up and running the tests */ + + void doForCurrentOffsets(BiConsumer consumer) + { - offsetConsumer.assign(partitions()); - partitions().forEach(tp -> consumer.accept(tp, offsetConsumer.position(tp))); - offsetConsumer.unsubscribe(); ++ partitions().forEach(tp -> ++ { ++ String partition = Integer.toString(tp.partition()); ++ Optional offset = partitionStatisticsRepository.findById(partition).map(document -> document.offset); ++ consumer.accept(tp, offset.orElse(0l)); ++ }); + } + + List partitions() + { + return + IntStream + .range(0, PARTITIONS) + .mapToObj(partition -> new TopicPartition(TOPIC, partition)) + .collect(Collectors.toList()); + } + + + void send100Messages(Function messageGenerator) + { + long i = 0; + + for (int partition = 0; partition < 10; partition++) + { + for (int key = 0; key < 10; key++) + { + Bytes value = messageGenerator.apply(++i); + + ProducerRecord record = + new ProducerRecord<>( + TOPIC, + partition, + Integer.toString(key%2), + value); + + kafkaProducer.send(record, (metadata, e) -> + { + if (metadata != null) + { + log.debug( + "{}|{} - {}={}", + metadata.partition(), + metadata.offset(), + record.key(), + record.value()); + } + else + { + log.warn( + "Exception for {}={}: {}", + record.key(), + record.value(), + e.toString()); + } + }); + } + } + } + + + @BeforeEach + public void init() + { + testHandler = record -> {} ; + + oldOffsets = new HashMap<>(); + newOffsets = new HashMap<>(); + receivedRecords = new HashSet<>(); + + doForCurrentOffsets((tp, offset) -> + { + oldOffsets.put(tp, offset - 1); + newOffsets.put(tp, offset - 1); + }); + + Consumer> captureOffsetAndExecuteTestHandler = + record -> + { + newOffsets.put( + new TopicPartition(record.topic(), record.partition()), + record.offset()); + receivedRecords.add(record); + testHandler.accept(record); + }; + + endlessConsumer = + new EndlessConsumer<>( + executor, + repository, + properties.getClientId(), + properties.getTopic(), + kafkaConsumer, + captureOffsetAndExecuteTestHandler); + + endlessConsumer.start(); + } + + @AfterEach + public void deinit() + { + try + { + endlessConsumer.stop(); + } + catch (Exception e) + { + log.info("Exception while stopping the consumer: {}", e.toString()); + } + } + + + @TestConfiguration + @Import(ApplicationConfiguration.class) + public static class Configuration + { + @Bean + Serializer serializer() + { + return new LongSerializer(); + } + + @Bean + KafkaProducer kafkaProducer(ApplicationProperties properties) + { + Properties props = new Properties(); + props.put("bootstrap.servers", properties.getBootstrapServer()); + props.put("linger.ms", 100); + props.put("key.serializer", StringSerializer.class.getName()); + props.put("value.serializer", BytesSerializer.class.getName()); + + return new KafkaProducer<>(props); + } - - @Bean - KafkaConsumer offsetConsumer(ApplicationProperties properties) - { - Properties props = new Properties(); - props.put("bootstrap.servers", properties.getBootstrapServer()); - props.put("client.id", "OFFSET-CONSUMER"); - props.put("group.id", properties.getGroupId()); - props.put("key.deserializer", BytesDeserializer.class.getName()); - props.put("value.deserializer", BytesDeserializer.class.getName()); - - return new KafkaConsumer<>(props); - } + } + }