最近有个需求,需要把我们Structured Streaming处理后的实时数据,发送到Redis一份。官网并没有提供redis输出方式。之前我们使用的是foreachBatch这种方式,可以同时输出到关系型数据库,kafka等,但是官方没提供输出方法的redis就有点难处理。后来看官方文档,官方推荐我们使用foreach进行输出。对于我们这种需要往多个数据源同时输出的情况,我们需要自定义Output Sink:
自定义sink需要继承自ForeachWriter。以下是我写的同时输出到kafka,redis和mysql的sink类
package xds.DataCleaning_201905
import java.sql.{Connection, PreparedStatement}
import java.util
import java.util.Date
import org.apache.kafka.clients.producer.{KafkaProducer, ProducerRecord}
import org.apache.spark.sql.{ForeachWriter, Row}
import org.json4s.jackson.JsonMethods.{compact, render}
import redis.clients.jedis.{Jedis}
import xds.Utils.{DateUtils, KafkaProducerUtils, MysqlManager, RedisClient}
import org.json4s._
import org.json4s.JsonDSL._
import org.json4s.jackson.JsonMethods._
/**
* wh 20190621
*
*
* 对于partition_id的每个分区:
*
* 对于epoch_id的流数据的每个批次/纪元:
*
* 方法open(partitionId,epochId)被调用。
*
* 如果open(...)返回true,则对于分区和批处理/纪元中的每一行,将调用方法进程(行)。
*
* 调用方法close(错误),在处理行时看到错误(如果有)。
*/
class MySink extends ForeachWriter [Row]{
val kafkaTopic : String = "LS_VD_CL"
var jedis: Jedis = _
var connection :Connection = _
var statementToInsert : PreparedStatement = _
var kafkaProducer : KafkaProducer[String, String] = _
override def open(partitionId: Long, version: Long): Boolean = {
jedis = RedisClient.pool.getResource
connection = MysqlManager.getMysqlManager.getConnection
kafkaProducer = KafkaProducerUtils.getProducer
connection.setAutoCommit(false)
statementToInsert = connection.prepareStatement("insert into t_videodata_1min (CreateTime,VehicleCount,Speed,ID_Link,ID_Station,ID_Lane,ID_TrafficSource,Type)" +
"values (?,?,?,?,?,?,?,?)")
println("open connection !")
true
}
override def process(value: Row): Unit = {
//获取row中每一个字段
val CreateTime:Date = value.getAs[Date](0)
val VehicleCount:Float = value.getAs[Float](1)
val Speed:Float= value.getAs[Float](2)
val ID_Link:String = value.getAs[String](3)
val ID_Station:String = value.getAs[String](4)
val ID_Lane:String = value.getAs[String](5)
val ID_TrafficSource:String = value.getAs[String](6)
val Type:Integer = value.getAs[Integer](7)
//以下为存入redis
val map :util.HashMap[String,String]= new util.HashMap[String,String]
map.put("VehicleCount",VehicleCount.toString)
map.put("Speed",Speed.toString)
map.put("ID_Lane",ID_Lane)
val hourMin = DateUtils.dateToStr(CreateTime,"HHmm")
jedis.hmset("C"+hourMin+ID_Link+"#"+ID_TrafficSource,map)
val createTimeStr = DateUtils.dateToStr(CreateTime,"yyyy-MM-dd HH:mm:ss")
//以下为存入mysql
statementToInsert.setObject(1,CreateTime)
statementToInsert.setObject(2,VehicleCount)
statementToInsert.setObject(3,Speed)
statementToInsert.setObject(4,ID_Link)
statementToInsert.setObject(5,ID_Station)
statementToInsert.setObject(6,ID_Lane)
statementToInsert.setObject(7,ID_TrafficSource)
statementToInsert.setObject(8,Type)
statementToInsert.addBatch()
//以下为发至kafka
val messageToKafka = ("ID_TrafficSource" -> ID_TrafficSource) ~
("CreateTime" -> createTimeStr)~
("ID_Station" -> ID_Station) ~
("ID_Link" -> ID_Link)~
("ID_Lane" -> ID_Lane)~
("VehicleCount" -> VehicleCount) ~
("Speed" -> Speed) ~
("Type" -> Type.toString)
val jsonToKafka = compact(render(messageToKafka))//封装成json
kafkaProducer.send(new ProducerRecord(kafkaTopic,jsonToKafka))
}
//记得关闭各种连接
override def close(errorOrNull: Throwable): Unit = {
//关闭连接
println("close connection !")
statementToInsert.executeBatch() //批量执行
connection.commit //提交
//注意关闭各种连接
statementToInsert.close()
connection.close()
jedis.close()
}
}
主函数里面我们只需要如下调用即可:
val query = df.writeStream.outputMode("append").foreach(new MySink).start()
query.awaitTermination()