spark streaming窗口聚合操作后如何管理offset

spark streaming窗口聚合操作后如何管理offset
很多知识星球球友问过浪尖一个问题:	
	
就是spark streaming经过窗口的聚合操作之后,再去管理offset呢?


对于spark streaming来说窗口操作之后,是无法管理offset的,因为offset的存储于HasOffsetRanges。只有kafkaRDD继承了他,所以假如我们对KafkaRDD进行了转化之后就无法再获取offset了。


还有窗口之后的offset的管理,也是很麻烦的,主要原因就是窗口操作会包含若干批次的RDD数据,那么提交offset我们只需要提交最近的那个批次的kafkaRDD的offset即可。如何获取呢?


对于spark 来说代码执行位置分为driver和executor,我们希望再driver端获取到offset,在处理完结果提交offset,或者直接与结果一起管理offset。


说到driver端执行,其实我们只需要使用transform获取到offset信息,然后在输出操作foreachrdd里面使用提交即可。

package bigdata.spark.SparkStreaming.kafka010	
	
import java.util.Properties	
	
import org.apache.kafka.clients.consumer.{Consumer, ConsumerRecord, KafkaConsumer}	
import org.apache.kafka.common.TopicPartition	
import org.apache.kafka.common.serialization.StringDeserializer	
import org.apache.spark.rdd.RDD	
import org.apache.spark.streaming.kafka010._	
import org.apache.spark.streaming.{Seconds, StreamingContext}	
import org.apache.spark.{SparkConf, TaskContext}	
	
import scala.collection.JavaConverters._	
import scala.collection.mutable	
	
object kafka010NamedRDD {	
   def main(args: Array[String]) {	
      //    创建一个批处理时间是2s的context 要增加环境变量	
      val sparkConf = new SparkConf().setAppName("DirectKafkaWordCount").setMaster("local[*]")	
      val ssc = new StreamingContext(sparkConf, Seconds(5))	
	
     ssc.checkpoint("/opt/checkpoint")	
	
      //    使用broker和topic创建DirectStream	
      val topicsSet = "test".split(",").toSet	
      val kafkaParams = Map[String, Object]("bootstrap.servers" -> "mt-mdh.local:9093",	
        "key.deserializer"->classOf[StringDeserializer],	
        "value.deserializer"-> classOf[StringDeserializer],	
        "group.id"->"test4",	
        "auto.offset.reset" -> "latest",	
        "enable.auto.commit"->(false: java.lang.Boolean))	
	
     // 没有接口提供 offset	
      val messages = KafkaUtils.createDirectStream[String, String](	
        ssc,	
        LocationStrategies.PreferConsistent,	
        ConsumerStrategies.Subscribe[String, String](topicsSet, kafkaParams,getLastOffsets(kafkaParams ,topicsSet)))//	
     var A:mutable.HashMap[String,Array[OffsetRange]] = new mutable.HashMap()	
	
     val trans = messages.transform(r =>{	
       val offsetRanges = r.asInstanceOf[HasOffsetRanges].offsetRanges	
       A += ("rdd1"->offsetRanges)	
       r	
     }).countByWindow(Seconds(10), Seconds(5))	
     trans.foreachRDD(rdd=>{	
	
       if(!rdd.isEmpty()){	
         val offsetRanges = A.get("rdd1").get//.asInstanceOf[HasOffsetRanges].offsetRanges	
	
         rdd.foreachPartition { iter =>	
           val o: OffsetRange = offsetRanges(TaskContext.get.partitionId)	
           println(s"${o.topic} ${o.partition} ${o.fromOffset} ${o.untilOffset}")	
         }	
	
         println(rdd.count())	
         println(offsetRanges)	
         // 手动提交offset ,前提是禁止自动提交	
         messages.asInstanceOf[CanCommitOffsets].commitAsync(offsetRanges)	
	
       }	
//       A.-("rdd1")	
     })	
      //    启动流	
      ssc.start()	
      ssc.awaitTermination()	
    }	
  def getLastOffsets(kafkaParams : Map[String, Object],topics:Set[String]): Map[TopicPartition, Long] ={	
    val props = new Properties()	
    props.putAll(kafkaParams.asJava)	
    val consumer = new KafkaConsumer[String, String](props)	
    consumer.subscribe(topics.asJavaCollection)	
    paranoidPoll(consumer)	
    val map = consumer.assignment().asScala.map { tp =>	
      println(tp+"---" +consumer.position(tp))	
      tp -> (consumer.position(tp))	
    }.toMap	
    println(map)	
    consumer.close()	
    map	
  }	
  def paranoidPoll(c: Consumer[String, String]): Unit = {	
    val msgs = c.poll(0)	
    if (!msgs.isEmpty) {	
      // position should be minimum offset per topicpartition	
      msgs.asScala.foldLeft(Map[TopicPartition, Long]()) { (acc, m) =>	
        val tp = new TopicPartition(m.topic, m.partition)	
        val off = acc.get(tp).map(o => Math.min(o, m.offset)).getOrElse(m.offset)	
        acc + (tp -> off)	
      }.foreach { case (tp, off) =>	
        c.seek(tp, off)	
      }	
    }	
  }	
}	

更多flink,spark的神奇操作,请加入浪尖知识星球,与800+好友一起学习。

www.zeeklog.com  - spark streaming窗口聚合操作后如何管理offset