我们从Spark的官方文档可以知道,维护Spark内部维护Kafka便宜了信息是存储在HasOffsetRanges类的offsetRanges中,我们可以在Spark Streaming程序里面获取这些信息:
val offsetsList = rdd.asInstanceOf[HasOffsetRanges].offsetRanges这样我们就可以获取所以分区消费信息,只需要遍历offsetsList,然后将这些信息发送到Zookeeper即可更新Kafka消费的偏移量。完整的代码片段如下:
val messages = KafkaUtils.createDirectStream[String, String, StringDecoder, StringDecoder](ssc, kafkaParams, topicsSet) messages.foreachRDD(rdd => { val offsetsList = rdd.asInstanceOf[HasOffsetRanges].offsetRanges val kc = new KafkaCluster(kafkaParams) for (offsets < - offsetsList) { val topicAndPartition = TopicAndPartition("iteblog", offsets.partition) val o = kc.setConsumerOffsets(args(0), Map((topicAndPartition, offsets.untilOffset))) if (o.isLeft) { println(s"Error updating the offset to Kafka cluster: ${o.left.get}") } } })
从图中我们可以看到KafkaOffsetMonitor监控软件已经可以监控到Kafka相关分区的消费情况,这对监控我们整个Spark Streaming程序来非常重要,因为我们可以任意时刻了解Spark读取速度。另外,KafkaCluster工具类的完整代码如下:
package org.apache.spark.streaming.kafka import kafka.api.OffsetCommitRequest import kafka.common.{ErrorMapping, OffsetMetadataAndError, TopicAndPartition} import kafka.consumer.SimpleConsumer import org.apache.spark.SparkException import org.apache.spark.streaming.kafka.KafkaCluster.SimpleConsumerConfig import scala.collection.mutable.ArrayBuffer import scala.util.Random import scala.util.control.NonFatal /** * User: 过往记忆 * Date: 2015-06-02 * Time: 下午23:46 * bolg: http://www.iteblog.com * 本文地址:http://www.iteblog.com/archives/1381 * 过往记忆博客,专注于hadoop、hive、spark、shark、flume的技术博客,大量的干货 * 过往记忆博客微信公共帐号:iteblog_hadoop */ class KafkaCluster(val kafkaParams: Map[String, String]) extends Serializable { type Err = ArrayBuffer[Throwable] @transient private var _config: SimpleConsumerConfig = null def config: SimpleConsumerConfig = this.synchronized { if (_config == null) { _config = SimpleConsumerConfig(kafkaParams) } _config } def setConsumerOffsets(groupId: String, offsets: Map[TopicAndPartition, Long] ): Either[Err, Map[TopicAndPartition, Short]] = { setConsumerOffsetMetadata(groupId, offsets.map { kv => kv._1 -> OffsetMetadataAndError(kv._2) }) } def setConsumerOffsetMetadata(groupId: String, metadata: Map[TopicAndPartition, OffsetMetadataAndError] ): Either[Err, Map[TopicAndPartition, Short]] = { var result = Map[TopicAndPartition, Short]() val req = OffsetCommitRequest(groupId, metadata) val errs = new Err val topicAndPartitions = metadata.keySet withBrokers(Random.shuffle(config.seedBrokers), errs) { consumer => val resp = consumer.commitOffsets(req) val respMap = resp.requestInfo val needed = topicAndPartitions.diff(result.keySet) needed.foreach { tp: TopicAndPartition => respMap.get(tp).foreach { err: Short => if (err == ErrorMapping.NoError) { result += tp -> err } else { errs.append(ErrorMapping.exceptionFor(err)) } } } if (result.keys.size == topicAndPartitions.size) { return Right(result) } } val missing = topicAndPartitions.diff(result.keySet) errs.append(new SparkException(s"Couldn't set offsets for ${missing}")) Left(errs) } private def withBrokers(brokers: Iterable[(String, Int)], errs: Err) (fn: SimpleConsumer => Any): Unit = { brokers.foreach { hp => var consumer: SimpleConsumer = null try { consumer = connect(hp._1, hp._2) fn(consumer) } catch { case NonFatal(e) => errs.append(e) } finally { if (consumer != null) { consumer.close() } } } } def connect(host: String, port: Int): SimpleConsumer = new SimpleConsumer(host, port, config.socketTimeoutMs, config.socketReceiveBufferBytes, config.clientId) }