From: Kai Moritz Date: Sat, 13 Aug 2022 09:15:54 +0000 (+0200) Subject: Verbesserungen aus 'rebalance-listener' nach 'counting-consumer' gemerged X-Git-Tag: sumup-requests---lvm-2-tage~5^2 X-Git-Url: https://juplo.de/gitweb/?a=commitdiff_plain;h=487ca1ac0c73b1c07547d5316d145ed46490ba9d;hp=-c;p=demos%2Fkafka%2Ftraining Verbesserungen aus 'rebalance-listener' nach 'counting-consumer' gemerged --- 487ca1ac0c73b1c07547d5316d145ed46490ba9d diff --combined src/main/java/de/juplo/kafka/ApplicationConfiguration.java index 0000000,7a0a8ad..bb219d0 mode 000000,100644..100644 --- a/src/main/java/de/juplo/kafka/ApplicationConfiguration.java +++ b/src/main/java/de/juplo/kafka/ApplicationConfiguration.java @@@ -1,0 -1,76 +1,64 @@@ + package de.juplo.kafka; + + 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; + + + @Configuration + @EnableConfigurationProperties(ApplicationProperties.class) + public class ApplicationConfiguration + { + @Bean + public KeyCountingRecordHandler keyCountingRecordHandler() + { + return new KeyCountingRecordHandler(); + } + - @Bean - public KeyCountingRebalanceListener keyCountingRebalanceListener( - KeyCountingRecordHandler keyCountingRecordHandler, - ApplicationProperties properties) - { - return new KeyCountingRebalanceListener( - keyCountingRecordHandler, - properties.getClientId()); - } - + @Bean + public EndlessConsumer endlessConsumer( + KafkaConsumer kafkaConsumer, + ExecutorService executor, - KeyCountingRebalanceListener keyCountingRebalanceListener, + KeyCountingRecordHandler keyCountingRecordHandler, + ApplicationProperties properties) + { + return + new EndlessConsumer<>( + executor, + properties.getClientId(), + properties.getTopic(), + kafkaConsumer, - keyCountingRebalanceListener, + keyCountingRecordHandler); + } + + @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("partition.assignment.strategy", "org.apache.kafka.clients.consumer.CooperativeStickyAssignor"); + props.put("group.id", properties.getGroupId()); + props.put("client.id", properties.getClientId()); + props.put("auto.offset.reset", properties.getAutoOffsetReset()); + props.put("auto.commit.interval.ms", (int)properties.getCommitInterval().toMillis()); + 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 --combined src/main/java/de/juplo/kafka/EndlessConsumer.java index 2310ccd,c7579b8..63a2f93 --- a/src/main/java/de/juplo/kafka/EndlessConsumer.java +++ b/src/main/java/de/juplo/kafka/EndlessConsumer.java @@@ -1,19 -1,15 +1,15 @@@ package de.juplo.kafka; + import lombok.RequiredArgsConstructor; import lombok.extern.slf4j.Slf4j; - import org.apache.kafka.clients.consumer.ConsumerRecord; - import org.apache.kafka.clients.consumer.ConsumerRecords; - import org.apache.kafka.clients.consumer.KafkaConsumer; + import org.apache.kafka.clients.consumer.*; + import org.apache.kafka.common.TopicPartition; + import org.apache.kafka.common.errors.RecordDeserializationException; import org.apache.kafka.common.errors.WakeupException; - import org.apache.kafka.common.serialization.StringDeserializer; import javax.annotation.PreDestroy; import java.time.Duration; - import java.util.Arrays; - import java.util.HashMap; - import java.util.Map; - import java.util.Optional; - import java.util.Properties; + import java.util.*; import java.util.concurrent.ExecutionException; import java.util.concurrent.ExecutorService; import java.util.concurrent.locks.Condition; @@@ -22,73 -18,41 +18,40 @@@ import java.util.concurrent.locks.Reent @Slf4j - public class EndlessConsumer implements Runnable + @RequiredArgsConstructor + public class EndlessConsumer implements Runnable { private final ExecutorService executor; - private final String bootstrapServer; - private final String groupId; private final String id; private final String topic; - private final String autoOffsetReset; + private final Consumer consumer; - private final ConsumerRebalanceListener consumerRebalanceListener; + private final RecordHandler handler; private final Lock lock = new ReentrantLock(); private final Condition condition = lock.newCondition(); private boolean running = false; private Exception exception; private long consumed = 0; - private KafkaConsumer consumer = null; - private Map> seen; - - - public EndlessConsumer( - ExecutorService executor, - String bootstrapServer, - String groupId, - String clientId, - String topic, - String autoOffsetReset) - { - this.executor = executor; - this.bootstrapServer = bootstrapServer; - this.groupId = groupId; - this.id = clientId; - this.topic = topic; - this.autoOffsetReset = autoOffsetReset; - } @Override public void run() { try { - Properties props = new Properties(); - props.put("bootstrap.servers", bootstrapServer); - props.put("group.id", groupId); - props.put("client.id", id); - props.put("auto.offset.reset", autoOffsetReset); - props.put("metadata.max.age.ms", "1000"); - props.put("key.deserializer", StringDeserializer.class.getName()); - props.put("value.deserializer", StringDeserializer.class.getName()); - - this.consumer = new KafkaConsumer<>(props); - log.info("{} - Subscribing to topic {}", id, topic); - consumer.subscribe(Arrays.asList(topic), consumerRebalanceListener); + consumer.subscribe(Arrays.asList(topic)); - seen = new HashMap<>(); - while (true) { - ConsumerRecords records = + ConsumerRecords records = consumer.poll(Duration.ofSeconds(1)); // Do something with the data... log.info("{} - Received {} messages", id, records.count()); - for (ConsumerRecord record : records) + for (ConsumerRecord record : records) { - consumed++; log.info( "{} - {}: {}/{} - {}={}", id, @@@ -99,28 -63,32 +62,32 @@@ record.value() ); - Integer partition = record.partition(); - String key = record.key() == null ? "NULL" : record.key(); - - if (!seen.containsKey(partition)) - seen.put(partition, new HashMap<>()); - - Map byKey = seen.get(partition); - - if (!byKey.containsKey(key)) - byKey.put(key, 0); + handler.accept(record); - int seenByKey = byKey.get(key); - seenByKey++; - byKey.put(key, seenByKey); + consumed++; } } } 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); @@@ -128,33 -96,10 +95,10 @@@ } finally { - log.info("{} - Closing the KafkaConsumer", id); - consumer.close(); - - for (Integer partition : seen.keySet()) - { - Map byKey = seen.get(partition); - for (String key : byKey.keySet()) - { - log.info( - "{} - Seen {} messages for partition={}|key={}", - id, - byKey.get(key), - partition, - key); - } - } - seen = null; - log.info("{} - Consumer-Thread exiting", id); } } - public Map> getSeen() - { - return seen; - } - private void shutdown() { shutdown(null); @@@ -165,9 -110,25 +109,25 @@@ lock.lock(); try { - running = false; - exception = e; - condition.signal(); + try + { + log.info("{} - Unsubscribing from topic {}", id, topic); + consumer.unsubscribe(); + } + catch (Exception ue) + { + log.error( + "{} - Error while unsubscribing from topic {}: {}", + id, + topic, + ue.toString()); + } + finally + { + running = false; + exception = e; + condition.signal(); + } } finally { @@@ -194,7 -155,7 +154,7 @@@ } } - public synchronized void stop() throws ExecutionException, InterruptedException + public synchronized void stop() throws InterruptedException { lock.lock(); try @@@ -217,22 -178,7 +177,7 @@@ public void destroy() throws ExecutionException, InterruptedException { log.info("{} - Destroy!", id); - try - { - stop(); - } - catch (IllegalStateException e) - { - log.info("{} - Was already stopped", id); - } - catch (Exception e) - { - log.error("{} - Unexpected exception while trying to stop the consumer", id, e); - } - finally - { - log.info("{}: Consumed {} messages in total, exiting!", id, consumed); - } + log.info("{}: Consumed {} messages in total, exiting!", id, consumed); } public boolean running() diff --combined src/main/java/de/juplo/kafka/KeyCountingRecordHandler.java index 0000000,099dcf7..83b3ff2 mode 000000,100644..100644 --- a/src/main/java/de/juplo/kafka/KeyCountingRecordHandler.java +++ b/src/main/java/de/juplo/kafka/KeyCountingRecordHandler.java @@@ -1,0 -1,46 +1,40 @@@ + package de.juplo.kafka; + + import lombok.extern.slf4j.Slf4j; + import org.apache.kafka.clients.consumer.ConsumerRecord; + + import java.util.HashMap; + import java.util.Map; + + + @Slf4j + public class KeyCountingRecordHandler implements RecordHandler + { + private final Map> seen = new HashMap<>(); + + + @Override + public void accept(ConsumerRecord record) + { + Integer partition = record.partition(); + String key = record.key() == null ? "NULL" : record.key().toString(); ++ ++ if (!seen.containsKey(partition)) ++ seen.put(partition, new HashMap<>()); ++ + Map byKey = seen.get(partition); + + if (!byKey.containsKey(key)) + byKey.put(key, 0l); + + long seenByKey = byKey.get(key); + seenByKey++; + byKey.put(key, seenByKey); + } + - public void addPartition(Integer partition, Map statistics) - { - seen.put(partition, statistics); - } - - public Map removePartition(Integer partition) - { - return seen.remove(partition); - } - + + public Map> getSeen() + { + return seen; + } + } diff --combined src/test/java/de/juplo/kafka/ApplicationTests.java index 0000000,5b13b7d..0909f2c mode 000000,100644..100644 --- a/src/test/java/de/juplo/kafka/ApplicationTests.java +++ b/src/test/java/de/juplo/kafka/ApplicationTests.java @@@ -1,0 -1,345 +1,342 @@@ + 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.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.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, + "consumer.commit-interval=1s" }) + @EmbeddedKafka(topics = TOPIC, partitions = PARTITIONS) + @EnableAutoConfiguration + @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; + @Autowired + ApplicationProperties properties; + @Autowired + ExecutorService executor; + @Autowired - KeyCountingRebalanceListener keyCountingRebalanceListener; - @Autowired + KeyCountingRecordHandler keyCountingRecordHandler; + + EndlessConsumer endlessConsumer; + Map oldOffsets; + Map newOffsets; + Set> receivedRecords; + + + /** Tests methods */ + + @Test + void commitsCurrentOffsetsOnSuccess() throws ExecutionException, InterruptedException + { + send100Messages((partition, key, counter) -> + { + Bytes value = new Bytes(valueSerializer.serialize(TOPIC, counter)); + return new ProducerRecord<>(TOPIC, partition, key, value); + }); + + await("100 records received") + .atMost(Duration.ofSeconds(30)) + .pollInterval(Duration.ofSeconds(1)) + .until(() -> receivedRecords.size() >= 100); + + await("Offsets committed") + .atMost(Duration.ofSeconds(10)) + .pollInterval(Duration.ofSeconds(1)) + .untilAsserted(() -> + { + checkSeenOffsetsForProgress(); + compareToCommitedOffsets(newOffsets); + }); + + assertThatExceptionOfType(IllegalStateException.class) + .isThrownBy(() -> endlessConsumer.exitStatus()) + .describedAs("Consumer should still be running"); + } + + @Test + void commitsOffsetOfErrorForReprocessingOnDeserializationError() + { + send100Messages((partition, key, counter) -> + { + Bytes value = counter == 77 + ? new Bytes(stringSerializer.serialize(TOPIC, "BOOM!")) + : new Bytes(valueSerializer.serialize(TOPIC, counter)); + return new ProducerRecord<>(TOPIC, partition, key, value); + }); + + await("Consumer failed") + .atMost(Duration.ofSeconds(30)) + .pollInterval(Duration.ofSeconds(1)) + .until(() -> !endlessConsumer.running()); + + checkSeenOffsetsForProgress(); + compareToCommitedOffsets(newOffsets); + + endlessConsumer.start(); + await("Consumer failed") + .atMost(Duration.ofSeconds(30)) + .pollInterval(Duration.ofSeconds(1)) + .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) + 1; + Long newOffset = newOffsets.get(tp) + 1; + 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 seekToEnd() + { + offsetConsumer.assign(partitions()); + offsetConsumer.seekToEnd(partitions()); + partitions().forEach(tp -> + { + // seekToEnd() works lazily: it only takes effect on poll()/position() + Long offset = offsetConsumer.position(tp); + log.info("New position for {}: {}", tp, offset); + }); + // The new positions must be commited! + offsetConsumer.commitSync(); + offsetConsumer.unsubscribe(); + } + + void doForCurrentOffsets(BiConsumer consumer) + { + offsetConsumer.assign(partitions()); + partitions().forEach(tp -> consumer.accept(tp, offsetConsumer.position(tp))); + offsetConsumer.unsubscribe(); + } + + List partitions() + { + return + IntStream + .range(0, PARTITIONS) + .mapToObj(partition -> new TopicPartition(TOPIC, partition)) + .collect(Collectors.toList()); + } + + + public interface RecordGenerator + { + public ProducerRecord generate(int partition, String key, long counter); + } + + void send100Messages(RecordGenerator recordGenerator) + { + long i = 0; + + for (int partition = 0; partition < 10; partition++) + { + for (int key = 0; key < 10; key++) + { + ProducerRecord record = + recordGenerator.generate(partition, Integer.toString(partition*10+key%2), ++i); + + 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() + { + seekToEnd(); + + oldOffsets = new HashMap<>(); + newOffsets = new HashMap<>(); + receivedRecords = new HashSet<>(); + + doForCurrentOffsets((tp, offset) -> + { + oldOffsets.put(tp, offset - 1); + newOffsets.put(tp, offset - 1); + }); + + TestRecordHandler captureOffsetAndExecuteTestHandler = + new TestRecordHandler(keyCountingRecordHandler) { + @Override + public void onNewRecord(ConsumerRecord record) + { + newOffsets.put( + new TopicPartition(record.topic(), record.partition()), + record.offset()); + receivedRecords.add(record); + } + }; + + endlessConsumer = + new EndlessConsumer<>( + executor, + properties.getClientId(), + properties.getTopic(), + kafkaConsumer, - keyCountingRebalanceListener, + 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); + } + } + }