drill-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (DRILL-4779) Kafka storage plugin support
Date Thu, 09 Nov 2017 21:15:08 GMT

    [ https://issues.apache.org/jira/browse/DRILL-4779?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16246553#comment-16246553
] 

ASF GitHub Bot commented on DRILL-4779:
---------------------------------------

Github user paul-rogers commented on a diff in the pull request:

    https://github.com/apache/drill/pull/1027#discussion_r150086292
  
    --- Diff: contrib/storage-kafka/src/main/java/org/apache/drill/exec/store/kafka/KafkaRecordReader.java
---
    @@ -0,0 +1,178 @@
    +/**
    + * Licensed to the Apache Software Foundation (ASF) under one
    + * or more contributor license agreements.  See the NOTICE file
    + * distributed with this work for additional information
    + * regarding copyright ownership.  The ASF licenses this file
    + * to you under the Apache License, Version 2.0 (the
    + * "License"); you may not use this file except in compliance
    + * with the License.  You may obtain a copy of the License at
    + *
    + * http://www.apache.org/licenses/LICENSE-2.0
    + *
    + * Unless required by applicable law or agreed to in writing, software
    + * distributed under the License is distributed on an "AS IS" BASIS,
    + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    + * See the License for the specific language governing permissions and
    + * limitations under the License.
    + */
    +package org.apache.drill.exec.store.kafka;
    +
    +import static org.apache.drill.exec.store.kafka.DrillKafkaConfig.DRILL_KAFKA_POLL_TIMEOUT;
    +
    +import java.util.Collection;
    +import java.util.Iterator;
    +import java.util.List;
    +import java.util.Set;
    +import java.util.concurrent.TimeUnit;
    +
    +import org.apache.drill.common.exceptions.ExecutionSetupException;
    +import org.apache.drill.common.expression.SchemaPath;
    +import org.apache.drill.exec.ExecConstants;
    +import org.apache.drill.exec.ops.FragmentContext;
    +import org.apache.drill.exec.ops.OperatorContext;
    +import org.apache.drill.exec.physical.impl.OutputMutator;
    +import org.apache.drill.exec.store.AbstractRecordReader;
    +import org.apache.drill.exec.store.kafka.KafkaSubScan.KafkaSubScanSpec;
    +import org.apache.drill.exec.store.kafka.decoders.MessageReader;
    +import org.apache.drill.exec.store.kafka.decoders.MessageReaderFactory;
    +import org.apache.drill.exec.util.Utilities;
    +import org.apache.drill.exec.vector.complex.impl.VectorContainerWriter;
    +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.common.TopicPartition;
    +import org.slf4j.Logger;
    +import org.slf4j.LoggerFactory;
    +
    +import com.google.common.base.Stopwatch;
    +import com.google.common.collect.Lists;
    +import com.google.common.collect.Sets;
    +public class KafkaRecordReader extends AbstractRecordReader {
    +  private static final Logger logger = LoggerFactory.getLogger(KafkaRecordReader.class);
    +  public static final long DEFAULT_MESSAGES_PER_BATCH = 4000;
    +
    +  private VectorContainerWriter writer;
    +  private MessageReader messageReader;
    +
    +  private boolean unionEnabled;
    +  private KafkaConsumer<byte[], byte[]> kafkaConsumer;
    +  private KafkaStoragePlugin plugin;
    +  private KafkaSubScanSpec subScanSpec;
    +  private long kafkaPollTimeOut;
    +  private long endOffset;
    +
    +  private long currentOffset;
    +  private long totalFetchTime = 0;
    +
    +  private List<TopicPartition> partitions;
    +  private final boolean enableAllTextMode;
    +  private final boolean readNumbersAsDouble;
    +
    +  private Iterator<ConsumerRecord<byte[], byte[]>> messageIter;
    +
    +  public KafkaRecordReader(KafkaSubScan.KafkaSubScanSpec subScanSpec, List<SchemaPath>
projectedColumns,
    +      FragmentContext context, KafkaStoragePlugin plugin) {
    +    setColumns(projectedColumns);
    +    this.enableAllTextMode = context.getOptions().getOption(ExecConstants.KAFKA_ALL_TEXT_MODE).bool_val;
    +    this.readNumbersAsDouble = context.getOptions()
    +        .getOption(ExecConstants.KAFKA_READER_READ_NUMBERS_AS_DOUBLE).bool_val;
    +    this.unionEnabled = context.getOptions().getOption(ExecConstants.ENABLE_UNION_TYPE);
    +    this.plugin = plugin;
    +    this.subScanSpec = subScanSpec;
    +    this.endOffset = subScanSpec.getEndOffset();
    +    this.kafkaPollTimeOut = Long.valueOf(plugin.getConfig().getDrillKafkaProps().getProperty(DRILL_KAFKA_POLL_TIMEOUT));
    +  }
    +
    +  @Override
    +  protected Collection<SchemaPath> transformColumns(Collection<SchemaPath>
projectedColumns) {
    +    Set<SchemaPath> transformed = Sets.newLinkedHashSet();
    +    if (!isStarQuery()) {
    +      for (SchemaPath column : projectedColumns) {
    +        transformed.add(column);
    +      }
    +    } else {
    +      transformed.add(Utilities.STAR_COLUMN);
    +    }
    +    return transformed;
    +  }
    +
    +  @Override
    +  public void setup(OperatorContext context, OutputMutator output) throws ExecutionSetupException
{
    +    this.writer = new VectorContainerWriter(output, unionEnabled);
    +    messageReader = MessageReaderFactory.getMessageReader(plugin.getConfig().getDrillKafkaProps());
    +    messageReader.init(context.getManagedBuffer(), Lists.newArrayList(getColumns()),
this.writer,
    +        this.enableAllTextMode, false, this.readNumbersAsDouble);
    +    kafkaConsumer = messageReader.getConsumer(plugin);
    +    partitions = Lists.newArrayListWithCapacity(1);
    +    TopicPartition topicPartition = new TopicPartition(subScanSpec.getTopicName(), subScanSpec.getPartitionId());
    +    partitions.add(topicPartition);
    +    kafkaConsumer.assign(partitions);
    +    currentOffset = subScanSpec.getStartOffset();
    +    kafkaConsumer.seek(topicPartition, currentOffset);
    +  }
    +
    +  private boolean hasNext() {
    +    if (currentOffset > endOffset) {
    +      logger.info("Read all the messages from Kafka. Start offset : {}, End offset :
{}", currentOffset, endOffset);
    +      return false;
    +    } else if (messageIter != null && messageIter.hasNext()) {
    +      return true;
    +    } else {
    +      return fetch();
    +    }
    +  }
    +
    +  private boolean fetch() {
    +    long startTime = System.currentTimeMillis();
    +    ConsumerRecords<byte[], byte[]> consumerRecords = kafkaConsumer.poll(kafkaPollTimeOut);
    +
    +    if (consumerRecords.isEmpty()) {
    +      logger.warn("Not able to fetch messages within {}milliseconds. Consider increasing
the value of {}",
    +          DRILL_KAFKA_POLL_TIMEOUT, kafkaPollTimeOut);
    +    }
    +
    +    long lastFetchTime = (System.currentTimeMillis() - startTime);
    +    logger.debug("Total number of messages fetched : " + consumerRecords.count());
    +    logger.debug("Time taken to fetch : " + (lastFetchTime / 1000) + " seconds");
    +    totalFetchTime += lastFetchTime;
    +
    +    messageIter = consumerRecords.iterator();
    +    return messageIter.hasNext();
    +  }
    +
    +  /**
    +   * KafkaConsumer.poll will fetch 500 messages per poll call. So hasNext will
    +   * take care of polling multiple times for this given batch next invocation
    +   */
    +  @Override
    +  public int next() {
    +    writer.allocate();
    +    writer.reset();
    +    Stopwatch watch = Stopwatch.createStarted();
    +
    +    int messageCount = 0;
    +    while (hasNext()) {
    +      ConsumerRecord<byte[], byte[]> consumerRecord = messageIter.next();
    +      currentOffset = consumerRecord.offset();
    +      writer.setPosition(messageCount);
    +      messageReader.readMessage(consumerRecord);
    +      ++messageCount;
    +      if (messageCount >= DEFAULT_MESSAGES_PER_BATCH) {
    +        break;
    +      }
    --- End diff --
    
    Seems awkward. Do you get to control the number of messages per fetch? If so, can you
loop over chunks to fill the batch, avoiding the need for chunks to span batches?
    A sketch:
    ```
    while (messageCount < DEFAULT_MESSAGES_PER_BATCH) {
      int fetchcount = Math.min(500, DEFAULT_MESSAGES_PER_BATCH - messageCount);
      messageIter = readNextChunk();
      for (ConsumerRecord<byte[], byte[]> consumerRecord : messageIter) {
        // Process record
        messageCount++;
      }
    }
    ```
    
    That said, however, our eventual goal is to limit batch sizes based on memory use, not
just row count. At that point, the current design will work better as you'll need to carry
over the iterator from one batch to the next because you won't know, in general, how many
records a batch can contain.


> Kafka storage plugin support
> ----------------------------
>
>                 Key: DRILL-4779
>                 URL: https://issues.apache.org/jira/browse/DRILL-4779
>             Project: Apache Drill
>          Issue Type: New Feature
>          Components: Storage - Other
>    Affects Versions: 1.11.0
>            Reporter: B Anil Kumar
>            Assignee: B Anil Kumar
>              Labels: doc-impacting
>             Fix For: 1.12.0
>
>
> Implement Kafka storage plugin will enable the strong SQL support for Kafka.
> Initially implementation can target for supporting json and avro message types



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Mime
View raw message