Spark Structured Streaming框架之数据输出源详解

Posted 2020-10-05

　　Spark Structured streaming API支持的输出源有：Console、Memory、File和Foreach。其中Console在前两篇博文中已有详述，而Memory使用非常简单。本文着重介绍File和Foreach两种方式，并介绍如何在源码基本扩展新的输出方式。

1. File

　　Structured Streaming支持将数据以File形式保存起来，其中支持的文件格式有四种:json、text、csv和parquet。其使用方式也非常简单只需设置checkpointLocation和path即可。checkpointLocation是检查点保存的路径，而path是真实数据保存的路径。

如下所示的测试例子：

// Create DataFrame representing the stream of input lines from connection to host:port val lines = spark.readStream .format("socket") .option("host", host) .option("port", port) .load()   // Split the lines into words val words = lines.as[String].flatMap(_.split(" "))   // Generate running word count val wordCounts = words.groupBy("value").count()   // Start running the query that prints the running counts to the console val query = wordCounts.writeStream .format("json") .option("checkpointLocation","root/jar") .option("path","/root/jar") .start()

注意：

    File形式不能设置"compelete"模型，只能设置"Append"模型。由于Append模型不能有聚合操作，所以将数据保存到外部File时，不能有聚合操作。

2. Foreach

　　foreach输出方式只需要实现ForeachWriter抽象类，并实现三个方法，当Structured Streaming接收到数据就会执行其三个方法，如下的测试示例：

/* * 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. */   // scalastyle:off println package org.apache.spark.examples.sql.streaming   import org.apache.spark.sql.SparkSession   /** * Counts words in UTF8 encoded, ‘\n‘ delimited text received from the network. * * Usage: StructuredNetworkWordCount <hostname> <port> * <hostname> and <port> describe the TCP server that Structured Streaming * would connect to receive data. * * To run this on your local machine, you need to first run a Netcat server * `$ nc -lk 9999` * and then run the example * `$ bin/run-example sql.streaming.StructuredNetworkWordCount * localhost 9999` */ object StructuredNetworkWordCount { def main(args: Array[String]) { if (args.length < 2) { System.err.println("Usage: StructuredNetworkWordCount <hostname> <port>") System.exit(1) }   val host = args(0) val port = args(1).toInt   val spark = SparkSession .builder .appName("StructuredNetworkWordCount") .getOrCreate()   import spark.implicits._   // Create DataFrame representing the stream of input lines from connection to host:port val lines = spark.readStream .format("socket") .option("host", host) .option("port", port) .load()   // Start running the query that prints the running counts to the console val query = wordCounts.writeStream .outputMode("append") .foreach(new ForearchWriter[Row]{         override def open(partitionId:Long,version:Long):Boolean={             println("open")             return true         }         override def process(value:Row):Unit={             val spark = SparkSession.builder.getOrCreate()             val seq = value.mkString.split(" ")             val row = Row.fromSeq(seq)             val rowRDD:RDD[Row] = sparkContext.getOrCreate().parallelize[Row](Seq(row))                          val userSchema = new StructType().add("name","String").add("age","String")             val peopleDF = spark.createDataFrame(rowRDD,userSchema)             peopleDF.createOrReplaceTempView(myTable)             spark.sql("select * from myTable").show()         }                  override def close(errorOrNull:Throwable):Unit={             println("close")         }      }) .start()   query.awaitTermination() } } // scalastyle:on println

 

　　上述程序是直接继承ForeachWriter类的接口，并实现了open()、process()、close()三个方法。若采用显示定义一个类来实现，需要注意Scala的泛型设计，如下所示：

class myForeachWriter[T<:Row](stream:CatalogTable) extends ForearchWriter[T]{     override def open(partionId:Long,version:Long):Boolean={         println("open")         true     }          override def process(value:T):Unit={         println(value)     }          override def close(errorOrNull:Throwable):Unit={         println("close")     } }

 

3. 自定义

　　若上述Spark Structured Streaming API提供的数据输出源仍不能满足要求，那么还有一种方法可以使用：修改源码。

如下通过实现一种自定义的Console来介绍这种使用方式：

3.1 ConsoleSink

　　Spark有一个Sink接口，用户可以实现该接口的addBatch方法，其中的data参数是接收的数据，如下所示直接将其输出到控制台：

class ConsoleSink(streamName:String) extends Sink{     override def addBatch(batchId:Long, data;DataFrame):Unit = {         data.show()             } }

 

3.2 DataStreamWriter

　　在用户自定义的输出形式时，并调用start()方法后，Spark框架会去调用DataStreamWriter类的start()方法。所以用户可以直接在该方法中添加自定义的输出方式，如我们向其传递上述创建的ConsoleSink类示例，如下所示：

def start():StreamingQuery={     if(source == "memory"){         ...     }else if(source=="foreach"){         ...     }else if(source=="consoleSink"){         val streamName:String = extraOption.get("streamName") mathc{             case Some(str):str             case None=>throw new AnalysisException("streamName option must be specified for Sink")         }                  val sink = new consoleSink(streamName)         df.sparkSession.sessionState.streamingQueryManager.startQuery(             extraOption.get("queryName"),             extraOption.get("checkpointLocation"),             df,             sink,             outputMode,             useTempCheckpointLocaltion = true,             recoverFromCheckpointLocation = false,             trigger = trigger         )     }else{         ...     } }

3.3 Structured Streaming

　　在前两部修改和实现完成后，用户就可以按正常的Structured Streaming API方式使用了，唯一不同的是在输出形式传递的参数是"consoleSink"字符串，如下所示：

def execute(stream:CatalogTable):Unit={     val spark = SparkSession .builder .appName("StructuredNetworkWordCount") .getOrCreate()     /**1. 获取数据对象DataFrame*/     val lines = spark.readStream .format("socket") .option("host", "localhost") .option("port", 9999) .load()          /**2. 启动Streaming开始接受数据源的信息*/     val query:StreamingQuery = lines.writeStream                 .outputMode("append")                 .format("consoleSink")                 .option("streamName","myStream")                 .start()                      query.awaitTermination() }

4. 参考文献

[1]. Structured Streaming Programming Guide.

[2]. Kafka Integration Guide.