sparkstreaming窗口及聚合操作后怎么管理offset

private[spark] class KafkaRDD[K, V](    sc: SparkContext,    val kafkaParams: ju.Map[String, Object],    val offsetRanges: Array[OffsetRange],    val preferredHosts: ju.Map[TopicPartition, String],    useConsumerCache: Boolean) extends RDD[ConsumerRecord[K, V]](sc, Nil)  with Logging with HasOffsetRanges {

trait HasOffsetRanges {  def offsetRanges: Array[OffsetRange]}

var A:mutable.HashMap[String,Array[OffsetRange]] = new mutable.HashMap()
val offsetRanges = r.asInstanceOf[HasOffsetRanges].offsetRangesA += ("rdd1"->offsetRanges)

package bigdata.spark.SparkStreaming.kafka010
import java.util.Properties
import org.apache.kafka.clients.consumer.{Consumer, ConsumerRecord, KafkaConsumer}import org.apache.kafka.common.TopicPartitionimport org.apache.kafka.common.serialization.StringDeserializerimport org.apache.spark.rdd.RDDimport 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]) {      //    創(chuàng)建一個(gè)批處理時(shí)間是2s的context 要增加環(huán)境變量      val sparkConf = new SparkConf().setAppName("DirectKafkaWordCount").setMaster("local[*]")      val ssc = new StreamingContext(sparkConf, Seconds(5))
     ssc.checkpoint("/opt/checkpoint")
      //    使用broker和topic創(chuàng)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))
     // 沒(méi)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)         // 手動(dòng)提交offset ，前提是禁止自動(dòng)提交         messages.asInstanceOf[CanCommitOffsets].commitAsync(offsetRanges)
       }//       A.-("rdd1")     })      //    啟動(dòng)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)      }    }  }}