生产SparkStreaming数据零丢失最佳实践(含代码)

MySQL创建存储offset的表格

mysql> use test
mysql> create table hlw_offset(
        topic varchar(32),
        groupid varchar(50),
        partitions int,
        fromoffset bigint,
        untiloffset bigint,
        primary key(topic,groupid,partitions)
        );

Maven依赖包

<scala.version>2.11.8</scala.version>
<spark.version>2.3.1</spark.version>
<scalikejdbc.version>2.5.0</scalikejdbc.version>
--------------------------------------------------
<dependency>
    <groupId>org.scala-lang</groupId>
    <artifactId>scala-library</artifactId>
    <version>${scala.version}</version>
</dependency>
<dependency>
    <groupId>org.apache.spark</groupId>
    <artifactId>spark-core_2.11</artifactId>
    <version>${spark.version}</version>
</dependency>
<dependency>
    <groupId>org.apache.spark</groupId>
    <artifactId>spark-sql_2.11</artifactId>
    <version>${spark.version}</version>
</dependency>
<dependency>
    <groupId>org.apache.spark</groupId>
    <artifactId>spark-streaming_2.11</artifactId>
    <version>${spark.version}</version>
</dependency>
<dependency>
    <groupId>org.apache.spark</groupId>
    <artifactId>spark-streaming-kafka-0-8_2.11</artifactId>
    <version>${spark.version}</version>
</dependency>
<dependency>
    <groupId>mysql</groupId>
    <artifactId>mysql-connector-java</artifactId>
    <version>5.1.27</version>
</dependency>
<!-- https://mvnrepository.com/artifact/org.scalikejdbc/scalikejdbc -->
<dependency>
    <groupId>org.scalikejdbc</groupId>
    <artifactId>scalikejdbc_2.11</artifactId>
    <version>2.5.0</version>
</dependency>
<dependency>
    <groupId>org.scalikejdbc</groupId>
    <artifactId>scalikejdbc-config_2.11</artifactId>
    <version>2.5.0</version>
</dependency>
<dependency>
    <groupId>com.typesafe</groupId>
    <artifactId>config</artifactId>
    <version>1.3.0</version>
</dependency>
<dependency>
    <groupId>org.apache.commons</groupId>
    <artifactId>commons-lang3</artifactId>
    <version>3.5</version>
</dependency>

实现思路

1）StreamingContext
2）从kafka中获取数据(从外部存储获取offset-->根据offset获取kafka中的数据)
3）根据业务进行逻辑处理
4）将处理结果存到外部存储中--保存offset
5）启动程序，等待程序结束

代码实现

1. SparkStreaming主体代码如下

   import kafka.common.TopicAndPartition
   import kafka.message.MessageAndMetadata
   import kafka.serializer.StringDecoder
   import org.apache.spark.SparkConf
   import org.apache.spark.streaming.kafka.{HasOffsetRanges, KafkaUtils}
   import org.apache.spark.streaming.{Seconds, StreamingContext}
   import scalikejdbc._
   import scalikejdbc.config._
   object JDBCOffsetApp {
    def main(args: Array[String]): Unit = {
      //创建SparkStreaming入口
      val conf = new SparkConf().setMaster("local[2]").setAppName("JDBCOffsetApp")
      val ssc = new StreamingContext(conf,Seconds(5))
      //kafka消费主题
      val topics = ValueUtils.getStringValue("kafka.topics").split(",").toSet
      //kafka参数
      //这里应用了自定义的ValueUtils工具类，来获取application.conf里的参数，方便后期修改
      val kafkaParams = Map[String,String](
        "metadata.broker.list"->ValueUtils.getStringValue("metadata.broker.list"),
        "auto.offset.reset"->ValueUtils.getStringValue("auto.offset.reset"),
        "group.id"->ValueUtils.getStringValue("group.id")
      )
      //先使用scalikejdbc从MySQL数据库中读取offset信息
      //+------------+------------------+------------+------------+-------------+
      //| topic      | groupid          | partitions | fromoffset | untiloffset |
      //+------------+------------------+------------+------------+-------------+
      //MySQL表结构如上，将“topic”，“partitions”，“untiloffset”列读取出来
      //组成 fromOffsets: Map[TopicAndPartition, Long]，后面createDirectStream用到
      DBs.setup()
      val fromOffset = DB.readOnly( implicit session => {
        SQL("select * from hlw_offset").map(rs => {
          (TopicAndPartition(rs.string("topic"),rs.int("partitions")),rs.long("untiloffset"))
        }).list().apply()
      }).toMap
      //如果MySQL表中没有offset信息，就从0开始消费；如果有，就从已经存在的offset开始消费
        val messages = if (fromOffset.isEmpty) {
          println("从头开始消费...")
          KafkaUtils.createDirectStream[String,String,StringDecoder,StringDecoder](ssc,kafkaParams,topics)
        } else {
          println("从已存在记录开始消费...")
          val messageHandler = (mm:MessageAndMetadata[String,String]) => (mm.key(),mm.message())
          KafkaUtils.createDirectStream[String,String,StringDecoder,StringDecoder,(String,String)](ssc,kafkaParams,fromOffset,messageHandler)
        }
        messages.foreachRDD(rdd=>{
          if(!rdd.isEmpty()){
            //输出rdd的数据量
            println("数据统计记录为："+rdd.count())
            //官方案例给出的获得rdd offset信息的方法，offsetRanges是由一系列offsetRange组成的数组
   //          trait HasOffsetRanges {
   //            def offsetRanges: Array[OffsetRange]
   //          }
            val offsetRanges = rdd.asInstanceOf[HasOffsetRanges].offsetRanges
            offsetRanges.foreach(x => {
              //输出每次消费的主题，分区，开始偏移量和结束偏移量
              println(s"---${x.topic},${x.partition},${x.fromOffset},${x.untilOffset}---")
             //将最新的偏移量信息保存到MySQL表中
              DB.autoCommit( implicit session => {
                SQL("replace into hlw_offset(topic,groupid,partitions,fromoffset,untiloffset) values (?,?,?,?,?)")
              .bind(x.topic,ValueUtils.getStringValue("group.id"),x.partition,x.fromOffset,x.untilOffset)
                .update().apply()
              })
            })
          }
        })
      ssc.start()
      ssc.awaitTermination()
    }
   }

2. 自定义的ValueUtils工具类如下

   import com.typesafe.config.ConfigFactory
   import org.apache.commons.lang3.StringUtils
   object ValueUtils {
   val load = ConfigFactory.load()
    def getStringValue(key:String, defaultValue:String="") = {
   val value = load.getString(key)
      if(StringUtils.isNotEmpty(value)) {
        value
      } else {
        defaultValue
      }
    }
   }

3. application.conf内容如下

   metadata.broker.list = "192.168.137.251:9092"
   auto.offset.reset = "smallest"
   group.id = "hlw_offset_group"
   kafka.topics = "hlw_offset"
   serializer.class = "kafka.serializer.StringEncoder"
   request.required.acks = "1"
   # JDBC settings
   db.default.driver = "com.mysql.jdbc.Driver"
   db.default.url="jdbc:mysql://hadoop000:3306/test"
   db.default.user="root"
   db.default.password="123456"

4. 自定义kafka producer

   import java.util.{Date, Properties}
   import kafka.producer.{KeyedMessage, Producer, ProducerConfig}
   object KafkaProducer {
    def main(args: Array[String]): Unit = {
      val properties = new Properties()
      properties.put("serializer.class",ValueUtils.getStringValue("serializer.class"))
      properties.put("metadata.broker.list",ValueUtils.getStringValue("metadata.broker.list"))
      properties.put("request.required.acks",ValueUtils.getStringValue("request.required.acks"))
      val producerConfig = new ProducerConfig(properties)
      val producer = new Producer[String,String](producerConfig)
      val topic = ValueUtils.getStringValue("kafka.topics")
      //每次产生100条数据
      var i = 0
      for (i <- 1 to 100) {
        val runtimes = new Date().toString
       val messages = new KeyedMessage[String, String](topic,i+"","hlw: "+runtimes)
        producer.send(messages)
      }
      println("数据发送完毕...")
    }
   }

测试

1. 启动kafka服务，并创建主题

   [hadoop@hadoop000 bin]$ ./kafka-server-start.sh -daemon /home/hadoop/app/kafka_2.11-0.10.0.1/config/server.properties
   [hadoop@hadoop000 bin]$ ./kafka-topics.sh --list --zookeeper localhost:2181/kafka
   [hadoop@hadoop000 bin]$ ./kafka-topics.sh --create --zookeeper localhost:2181/kafka --replication-factor 1 --partitions 1 --topic hlw_offset

2. 测试前查看MySQL中offset表，刚开始是个空表

   mysql> select * from hlw_offset;
   Empty set (0.00 sec)

3. 通过kafka producer产生500条数据

4. 启动SparkStreaming程序

   //控制台输出结果：
   从头开始消费...
   数据统计记录为：500
   ---hlw_offset,0,0,500---

查看MySQL表，offset记录成功

mysql> select * from hlw_offset;
+------------+------------------+------------+------------+-------------+
| topic      | groupid          | partitions | fromoffset | untiloffset |
+------------+------------------+------------+------------+-------------+
| hlw_offset | hlw_offset_group |          0 |          0 |         500 |
+------------+------------------+------------+------------+-------------+

1. 关闭SparkStreaming程序，再使用kafka producer生产300条数据,再次启动spark程序（如果spark从500开始消费，说明成功读取了offset，做到了只读取一次语义）

   //控制台结果输出：
   从已存在记录开始消费...
   数据统计记录为：300
   ---hlw_offset,0,500,800---

2. 查看更新后的offset MySQL数据

   mysql> select * from hlw_offset;
   +------------+------------------+------------+------------+-------------+
   | topic      | groupid          | partitions | fromoffset | untiloffset |
   +------------+------------------+------------+------------+-------------+
   | hlw_offset | hlw_offset_group |          0 |        500 |         800 |
   +------------+------------------+------------+------------+-------------+