大數據面試·Spark篇（二）

Spark Streaming任務延遲監控及告警

1.需求

監控批次處理時間，若超過閾值則告警，每次告警間隔2分鐘

2.自定義StreamingListener

class SparkStreamingDelayListener(private val appName:String, private val duration: Int,private val times: Int) extends StreamingListener{

  private val logger = LoggerFactory.getLogger("SparkStreamingDelayListener")

//每個批次完成時執行
  override def onBatchCompleted(batchCompleted: StreamingListenerBatchCompleted): Unit = {
    val batchInfo = batchCompleted.batchInfo
    val processingStartTime = batchCompleted.batchInfo.processingStartTime
    val numRecords = batchCompleted.batchInfo.numRecords
    val processingEndTime = batchInfo.processingEndTime
    val processingDelay = batchInfo.processingDelay
    val totalDelay = batchInfo.totalDelay

    //將每次告警時間寫入redis,用以判斷告警間隔大於2分鐘
    val jedis = RedisClusterClient.getJedisClusterClient()
    val current_time = (System.currentTimeMillis / 1000).toInt
    val redis_time = jedis.get(appName)
    var flag = false
    if(redis_time==null || current_time-redis_time.toInt>120){
      jedis.set(appName,current_time.toString)
      flag = true
    }
    
    //若批次處理延遲大於批次時長指定倍數，並且告警間隔大約2分鐘，則告警
    if(totalDelay.get >= times * duration * 1000 && flag){
      val monitorContent = appName+": numRecords ->"+numRecords+",processingDelay ->"+processingDelay.get/1000+" s,totalDelay -> "+totalDelay.get/1000+"s"
      println(monitorContent)
      val msg = "Streaming_"+appName+"_DelayTime:"+totalDelay.get/1000+"S"
      val getURL = "http://node1:8002/message/weixin?msg="+msg
      HttpClient.doGet(getURL)
    }
  }
}

3.添加到streamingContext中

ssc.addStreamingListener(new SparkStreamingDelayListener("Userid2Redis", duration,times))