X-Git-Url: http://juplo.de/gitweb/?a=blobdiff_plain;f=src%2Fmain%2Fjava%2Fde%2Fjuplo%2Fkafka%2FEndlessConsumer.java;h=01f9057a055aa360bb8d98cca1641073df19afe0;hb=a6a0a22a5fa34a01b0e8b2bc1e0e2b82d7b60f33;hp=c2d44478c016551a491d350cd554699e483526c8;hpb=92882a04df8b54ae6ac000c14a9793f7411318f9;p=demos%2Fkafka%2Ftraining diff --git a/src/main/java/de/juplo/kafka/EndlessConsumer.java b/src/main/java/de/juplo/kafka/EndlessConsumer.java index c2d4447..01f9057 100644 --- a/src/main/java/de/juplo/kafka/EndlessConsumer.java +++ b/src/main/java/de/juplo/kafka/EndlessConsumer.java @@ -1,109 +1,98 @@ package de.juplo.kafka; +import lombok.RequiredArgsConstructor; import lombok.extern.slf4j.Slf4j; -import org.apache.kafka.clients.consumer.ConsumerRebalanceListener; -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.Clock; import java.time.Duration; +import java.time.Instant; import java.util.*; import java.util.concurrent.ExecutionException; import java.util.concurrent.ExecutorService; import java.util.concurrent.locks.Condition; import java.util.concurrent.locks.Lock; import java.util.concurrent.locks.ReentrantLock; +import java.util.regex.Pattern; @Slf4j -public class EndlessConsumer implements Runnable +@RequiredArgsConstructor +public class EndlessConsumer implements ConsumerRebalanceListener, Runnable { + final static Pattern PATTERN = Pattern.compile("\\W+"); + + private final ExecutorService executor; - private final String bootstrapServer; - private final String groupId; + private final PartitionStatisticsRepository repository; private final String id; private final String topic; - private final String autoOffsetReset; + private final Clock clock; + private final Duration commitInterval; + private final Consumer consumer; 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 final Map>> seen = 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); + log.info( + "{} - removing partition: {}, offset of next message {})", + id, + partition, + newOffset); + Map> removed = seen.remove(partition); + repository.save(new StatisticsDocument(partition, removed, consumer.position(tp))); + }); + } - public EndlessConsumer( - ExecutorService executor, - String bootstrapServer, - String groupId, - String clientId, - String topic, - String autoOffsetReset) + @Override + public void onPartitionsAssigned(Collection partitions) { - this.executor = executor; - 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); + StatisticsDocument document = + repository + .findById(Integer.toString(partition)) + .orElse(new StatisticsDocument(partition)); + if (document.offset >= 0) + { + // Only seek, if a stored offset was found + // Otherwise: Use initial offset, generated by Kafka + consumer.seek(tp, document.offset); + } + seen.put(partition, document.statistics); + }); } + @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), new ConsumerRebalanceListener() - { - @Override - public void onPartitionsRevoked(Collection partitions) - { - partitions.forEach(tp -> - { - log.info("{} - removing partition: {}", id, tp); - Map removed = seen.remove(tp.partition()); - for (String key : removed.keySet()) - { - log.info( - "{} - Seen {} messages for partition={}|key={}", - id, - removed.get(key), - tp.partition(), - key); - } - }); - } + consumer.subscribe(Arrays.asList(topic), this); - @Override - public void onPartitionsAssigned(Collection partitions) - { - partitions.forEach(tp -> - { - log.info("{} - adding partition: {}", id, tp); - seen.put(tp.partition(), new HashMap<>()); - }); - } - }); + Instant lastCommit = clock.instant(); while (true) { @@ -114,7 +103,6 @@ public class EndlessConsumer implements Runnable log.info("{} - Received {} messages", id, records.count()); for (ConsumerRecord record : records) { - consumed++; log.info( "{} - {}: {}/{} - {}={}", id, @@ -125,24 +113,64 @@ public class EndlessConsumer implements Runnable record.value() ); + consumed++; + Integer partition = record.partition(); - String key = record.key() == null ? "NULL" : record.key(); - Map byKey = seen.get(partition); + String user = record.key(); + Map> users = seen.get(partition); + + Map words = users.get(user); + if (words == null) + { + words = new HashMap<>(); + users.put(user, words); + } - if (!byKey.containsKey(key)) - byKey.put(key, 0); + for (String word : PATTERN.split(record.value())) + { + Long num = words.get(word); + if (num == null) + { + num = 1l; + } + else + { + num++; + } + words.put(word, num); + } + } - int seenByKey = byKey.get(key); - seenByKey++; - byKey.put(key, seenByKey); + if (lastCommit.plus(commitInterval).isBefore(clock.instant())) + { + log.debug("Storing data and offsets, last commit: {}", lastCommit); + seen.forEach((partiton, statistics) -> repository.save( + new StatisticsDocument( + partiton, + statistics, + consumer.position(new TopicPartition(topic, partiton))))); + lastCommit = clock.instant(); } } } catch(WakeupException e) { - log.info("{} - RIIING!", id); + log.info("{} - RIIING! Request to stop consumption - commiting current offsets!", id); 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()); + + shutdown(e); + } catch(Exception e) { log.error("{} - Unexpected error: {}", id, e.toString(), e); @@ -150,8 +178,6 @@ public class EndlessConsumer implements Runnable } finally { - log.info("{} - Closing the KafkaConsumer", id); - consumer.close(); log.info("{} - Consumer-Thread exiting", id); } } @@ -166,9 +192,25 @@ public class EndlessConsumer implements Runnable 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 { @@ -176,7 +218,7 @@ public class EndlessConsumer implements Runnable } } - public Map> getSeen() + public Map>> getSeen() { return seen; } @@ -200,7 +242,7 @@ public class EndlessConsumer implements Runnable } } - public synchronized void stop() throws ExecutionException, InterruptedException + public synchronized void stop() throws InterruptedException { lock.lock(); try @@ -223,22 +265,7 @@ public class EndlessConsumer implements Runnable 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()