开发环境
windows7 64、intellij idea 14.1.5、spark-1.5.2、scala 2.0.4、java1.7、maven3.05
将spark中的assembly包引入即可使用local模式运行相关的scala任务,注意不要使用scala2.11,非要使用的话先用这个版本的scala编译一遍spark哈
代码部分
pom文件
先附上pom.xml中的jar包依赖部分
<dependencies>
<dependency>
<groupId>org.scala-lang</groupId>
<artifactId>scala-library</artifactId>
<version>2.10.4</version>
</dependency>
<dependency>
<groupId>org.scala-lang</groupId>
<artifactId>scala-compiler</artifactId>
<version>2.10.4</version>
</dependency>
<dependency>
<groupId>org.scala-lang</groupId>
<artifactId>scala-reflect</artifactId>
<version>2.10.4</version>
</dependency>
<dependency>
<groupId>log4j</groupId>
<artifactId>log4j</artifactId>
<version>1.2.12</version>
</dependency>
<dependency>
<groupId>com.google.collections</groupId>
<artifactId>google-collections</artifactId>
<version>1.0</version>
</dependency>
<dependency>
<groupId>org.apache.spark</groupId>
<artifactId>spark-core_2.10</artifactId>
<version>1.5.2</version>
</dependency>
<dependency>
<groupId>mysql</groupId>
<artifactId>mysql-connector-java</artifactId>
<version>5.1.29</version>
</dependency>
<dependency>
<groupId>org.apache.spark</groupId>
<artifactId>spark-streaming_2.10</artifactId>
<version>1.5.2</version>
<scope>provided</scope>
</dependency>
<dependency>
<groupId>org.apache.spark</groupId>
<artifactId>spark-streaming-kafka_2.10</artifactId>
<version>1.5.2</version>
</dependency>
<dependency>
<groupId>com.jolbox</groupId>
<artifactId>bonecp</artifactId>
<version>0.8.0.RELEASE</version>
</dependency>
<dependency>
<groupId>postgresql</groupId>
<artifactId>postgresql</artifactId>
<version>9.2-1002.jdbc4</version>
</dependency>
</dependencies>
mysql连接池
连接池,利用boncp创建连接池供使用,引用自网上的代码,详细可看参考资料
/**
* Created by Administrator on 2016/2/29.
* 参考资料 http://www.myexception.cn/mysql/1934808.html
*/
import java.sql.{Connection, ResultSet}
import com.jolbox.bonecp.{BoneCP, BoneCPConfig}
import org.slf4j.LoggerFactory
object ConnectionPool {
val logger = LoggerFactory.getLogger(this.getClass)
private val connectionPool = {
try{
Class.forName("com.mysql.jdbc.Driver")
val config = new BoneCPConfig()
config.setJdbcUrl("jdbc:mysql://localhost:3306/test")
config.setUsername("etl")
config.setPassword("xxxxx")
config.setLazyInit(true)
config.setMinConnectionsPerPartition(3)
config.setMaxConnectionsPerPartition(5)
config.setPartitionCount(5)
config.setCloseConnectionWatch(true)
config.setLogStatementsEnabled(false)
Some(new BoneCP(config))
} catch {
case exception:Exception=>
logger.warn("Error in creation of connection pool"+exception.printStackTrace())
None
}
}
def getConnection:Option[Connection] ={
connectionPool match {
case Some(connPool) => Some(connPool.getConnection)
case None => None
}
}
def closeConnection(connection:Connection): Unit = {
if(!connection.isClosed) {
connection.close()
}
}
}
streaming处理程序
ssc程序,数据示例如下,删除了一些关键数据(不影响此次处理),__开头的key为系统自带属性.
__clientip=10.10.9.153&paymentstatus=0&__opip=&memberid=89385239&iamount=1&itype=16&oper_res=1&channeltype=8&__timestamp=1457252427&productid=112&selectbank=&icount=0&ordersrc=web&paymentip=61.159.104.134&orderdate=2016-03-06 16:19:55&subjecttype=zheanaiMessenger&oper_type=1&paydate=&orderamount=259.0&paymentchannel=16&oper_time=2016-03-06 16:20:27&orderid=127145727&iunit=month&bussinessid=80125727&isuse=0
__clientip=10.10.9.175&paymentstatus=0&__opip=&memberid=89378034&iamount=12&itype=17&oper_res=1&channeltype=75&__timestamp=1457252429&productid=124&selectbank=&icount=0&ordersrc=100&paymentip=59.37.137.119&orderdate=2016-03-06 16:20:29&subjecttype=zheanaiMessenger&oper_type=0&paydate=&orderamount=388.0&paymentchannel=1028&oper_time=2016-03-06 16:20:29&orderid=127145736&iunit=month&bussinessid=8012580&isuse=0
__clientip=10.10.9.153&paymentstatus=0&__opip=&memberid=75372899&iamount=12&itype=16&oper_res=1&channeltype=&__timestamp=1457252286&productid=131&selectbank=&icount=0&ordersrc=web&paymentip=113.226.244.206&orderdate=2016-03-06 16:18:06&subjecttype=zheanaiMessenger&oper_type=0&paydate=&orderamount=99.0&paymentchannel=307&oper_time=2016-03-06 16:18:06&orderid=127145700&iunit=month&bussinessid=80125477&isuse=0
__clientip=10.10.9.175&paymentstatus=0&__opip=&memberid=87634711&iamount=1&itype=16&oper_res=1&channeltype=8&__timestamp=1457252432&productid=129&selectbank=&icount=0&ordersrc=web&paymentip=114.246.35.251&orderdate=2016-03-06 16:19:05&subjecttype=zheanaiMessenger&oper_type=1&paydate=&orderamount=19.0&paymentchannel=16&oper_time=2016-03-06 16:20:32&orderid=127145713&iunit=month&bussinessid=66213022&isuse=0
__clientip=10.10.9.153&paymentstatus=0&__opip=&memberid=89172717&iamount=12&itype=17&oper_res=1&channeltype=77&__timestamp=1457252371&productid=124&selectbank=&icount=0&ordersrc=4&paymentip=111.126.43.83&orderdate=2016-03-06 16:19:31&subjecttype=zheanaiMessenger&oper_type=0&paydate=&orderamount=388.0&paymentchannel=1116&oper_time=2016-03-06 16:19:31&orderid=127145723&iunit=month&bussinessid=8012568&isuse=0
主要操作如下
读取,ssc自带的receiver
解析(valueSplit方法 处理成kv格式)
过滤filterRegex,类似sql中的where条件放弃一些不需要的数据,比如只需要买单的数据而不要下单数据
转换,getPlatform、getFormatDate,类似case when
创建了一个class命名为result,重写了toString方法。该class存放从kafka中处理后的所有需要的数据字段。
写入mysql,insertIntoMySQL,方法在每个partition中调用
另外代码中使用了getOrCreate以便恢复,利用了计数器简单统计了一下有效记录数
代码如下
import java.text.SimpleDateFormat
import java.util.Date
import com.zhenai.SqlConnection.ConnectionPool
import java.sql.Connection
import org.apache.log4j.PropertyConfigurator
import org.apache.spark.rdd.RDD
import org.apache.spark.streaming.kafka.KafkaUtils
import org.apache.spark.streaming.{Time,Seconds, StreamingContext}
import org.apache.spark.{SparkContext, SparkConf}
import org.joda.time.DateTime
import org.slf4j.LoggerFactory
import scala.collection.mutable.Map
/**
* Created by Administrator on 2016/2/25.
*/
object KafkaStreaming {
val logger = LoggerFactory.getLogger(this.getClass)
PropertyConfigurator.configure(System.getProperty("user.dir")+"\\src\\log4j.properties")
case class result(ftime:String,hour:String,orderid:Long,memberid:Long,platform:String,iamount:Double,orderamount:Double)extends Serializable{
override def toString: String="%s\t%s\t%d\t%d\t%s\t%.2f\t%.2f".format(ftime, hour,orderid,memberid,platform,iamount,orderamount)
}
def getFormatDate(date:Date,format:SimpleDateFormat): String ={
format.format(date)
}
def stringFormatTime(time:String,simpleformat:SimpleDateFormat): Date ={
simpleformat.parse(time)
}
// kafka中的value解析为Map
def valueSplit(value:String): Map[String,String] ={
val x = value.split("&")
val valueMap:Map[String,String] = Map()
x.foreach { kvs =>
if (!kvs.startsWith("__")){
val kv = kvs.split("=")
if (kv.length==2) {
valueMap += (kv(0) -> kv(1))
}
}
}
valueMap
}
// 实现类似where的条件,tips:优先过滤条件大的减少后续操作
def filterRegex(map:Map[String,String]): Boolean ={
//过滤操作类型,控制为支付操作
val oper_type = map.getOrElse("oper_type","-1")
if(!oper_type.equals("2") && !oper_type.equals("3"))
return false
// 过滤未支付成功记录
if(!map.getOrElse("paymentstatus","0").equals("1"))
return false
// 过滤无效支付ip
val paymentip = map.getOrElse("paymentip",null)
if (paymentip.startsWith("10.10")||paymentip.startsWith("183.62.134")||paymentip.contains("127.0.0.1"))
return false
return true
}
// 实现类似 case when的方法,上报的p字段不一定为数值
def getPlatform(p:String,x:Int): String ={
val platformname = (p,x) match{
case (p,x) if(Array[String]("1","2","3").contains(p)) => "wap"
case (p,x) if(Array[String]("4","8").contains(p)&& x!=18) =>"andriod"
case (p,x) if((Array[String]("5","7","51","100").contains(p))&&(p!=18)) => "ios"
case _ => "pc"
}
platformname
}
// 数据库写入
def insertIntoMySQL(con:Connection,sql:String,data:result): Unit ={
// println(data.toString)
try {
val ps = con.prepareStatement(sql)
ps.setString(1, data.ftime)
ps.setString(2, data.hour)
ps.setLong(3,data.orderid)
ps.setLong(4, data.memberid)
ps.setString(5, data.platform)
ps.setDouble(6, data.iamount)
ps.setDouble(7, data.orderamount)
ps.executeUpdate()
ps.close()
}catch{
case exception:Exception=>
logger.error("Error in execution of query "+exception.getMessage+"\n-----------------------\n"+exception.printStackTrace()+"\n-----------------------------")
}
}
def createContext(zkqurm:String,topic:scala.Predef.Map[String,Int],checkPointDir:String): StreamingContext ={
val simpleformat = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss")
val dateFormat = new SimpleDateFormat("yyyyMMdd")
val timeFormat = new SimpleDateFormat("HH:mm")
val sql ="insert into t_ssc_toufang_result_mi(ftime,hour,orderid,memberid,platform,iamount,orderamount) values(?,?,?,?,?,?,?);"
val conf = new SparkConf()
conf.setAppName("Scala Streaming read kafka")
// VM option -Dspark.master=local
// conf.setMaster("local[4]")
val sc = new SparkContext(conf)
val totalcounts = sc.accumulator(0L,"Total count")
val ssc = new StreamingContext(sc,Seconds(60))
//ssc.checkpoint(checkPointDir)
//统计各平台最近一分钟实时注册收入 时间段,平台,金额,订单数
val lines = KafkaUtils.createStream(ssc, zkqurm, "mytopic_local",topic).map(_._2)
val filterRecord = lines.filter(x => !x.isEmpty).map(valueSplit).filter(filterRegex).map{x =>
val orderdate = stringFormatTime(x.getOrElse("orderdate",null),simpleformat)
val day = getFormatDate(orderdate,dateFormat)
val hour = getFormatDate(orderdate,timeFormat)
var orderamount = x.getOrElse("orderamount","0").toDouble
if (x.getOrElse("oper_type",-1)==3)
orderamount = -1*orderamount
val res = new result(
day
,hour
,x.getOrElse("orderid",null).toLong
,x.getOrElse("memberid",null).toLong
,getPlatform(x.getOrElse("ordersrc",null),x.getOrElse("itype",null).toInt)
,x.getOrElse("iamount","0").toDouble
,orderamount
)
res
}
filterRecord.foreachRDD((x: RDD[result],time: Time) =>{
if(!x.isEmpty()) {
// 打印一下这一批batch的处理时间段以及累计的有效记录数(不含档次)
println("--"+new DateTime(time.milliseconds).toString("yyyy-MM-dd HH:mm:ss")+"--totalcounts:"+totalcounts.value+"-----")
x.foreachPartition{res =>
{
if(!res.isEmpty){
val connection = ConnectionPool.getConnection.getOrElse(null)
res.foreach {
r: result =>totalcounts.add(1L)
insertIntoMySQL(connection, sql, r)
}
ConnectionPool.closeConnection(connection)
}
}
}
}
})
ssc
}
// =================================================================
def main(args:Array[String]): Unit ={
val zkqurm = "10.10.10.177:2181,10.10.10.175:2181,10.10.10.179:2181"
val topic = scala.Predef.Map("t_fw_00015"->30)
val checkPointDir ="/user/root/sparkcheck"
val ssc = StreamingContext.getOrCreate(checkPointDir,
() => {
createContext(zkqurm, topic,checkPointDir)
})
ssc.start()
ssc.awaitTermination()
}
}
补充log4j.propertites文件代码
log4j.rootLogger=WARN,stdout,R
log4j.appender.stdout=org.apache.log4j.ConsoleAppender
log4j.appender.stdout.layout=org.apache.log4j.PatternLayout
log4j.appender.stdout.layout.ConversionPattern=%5p - %m%n
log4j.appender.R=org.apache.log4j.RollingFileAppender
log4j.appender.R.File=mapreduce_test.log
log4j.appender.R.MaxFileSize=10MB
log4j.appender.R.MaxBackupIndex=1
log4j.appender.R.layout=org.apache.log4j.PatternLayout
log4j.appender.R.layout.ConversionPattern=%p %t %c - %m%n
log4j.logger.com.codefutures=WARN
打包成jar运行
maven
由于是在idea intellij上配置的maven工程,直接使用maven打包了
build artifacts,操作流程如下
配置好要打包的类: file -> project structure -> artifacts,操作如下
选择main方法,点击确认,结果如下
发现结果包名为自己的项目名称,这里可以调整,选中jar一行右键重命名;引入了所有的类,选中全部右键remove掉,当然也可以根据需要选择保留需要的依赖,这里避免最终jar过大选择全部删掉
结果如下,注意有个输出路径,build完之后在这个路径下找到jar包
点击确认返回界面,左上的导航栏选择 build ->build artifacts ,选择刚刚配置的名字KafkaSsc:jar进行build
完成之后可到输出路径下发现有个jar包KafkaStreaming.jar,可以使用winrar打开看看里面都写了什么
运行
上传到linux下进行运行,指定两个jar包使用local模式,命令如下
spark-submit --master local[4] \
--class "com.xxxx.streaming.KafkaStreaming" \
--jars spark-streaming-kafka-assembly_2.10-1.5.0.jar,mysql-connector-java-5.1.29.jar \
KafkaStreaming.jar
运行情况如下,错误信息为log4j.properties未设置好,这个代码注释掉或者修改路径即可
一些个人想法
以上代码的不足
1、未考虑写入mysql失败的情况,处理方法:可以在insertIntoMySQL中做容错处理,这个不行我写到另一个地方存放起来,后台再起一个线程定时把这些数据写到目标mysql中去
2、代码可以更简洁一点、mysql、zk等参数应当支持配置文件处理
3、读取kafka数据建议KafkaUtils.createStream改为低阶api的实现KafkaUtils.createDirectStream去读取。具体可自行查询或者查看我的下一篇博客对这两个的一些总结
关于使用连接池说明:
1、不希望每写一条记录都创建一个连接,资源消耗大
2、分布式集群中,我们不知道这个数据记录会在那一台机,但是可以知道的是至少每一个partition里所有数据都是在一台机的
3、针对每个parition创建连接相对来说也很耗费资源,在处理时间段内整个ssc是大量sc的job组成的对rdd处理队列,存在多个mysql的长连接是必要的
为何不返回到driver端来执行呢?
个人认为在数据量不大的情况下这是可行的,完全可以返回所有的数据,并且批量写入mysql;但是如果数据量大的话很影响效率
关于实时业务场景的使用
细心的同学可能已经发现,在上面的ssc程序中并没有使用reduce之类的聚合操作,这个其实关系到一个业务场景,这个数据出来主要是用于报表的;首先这个数据量实在是很小,每分钟有效的不超过10条记录(实际上可以直接mysql搞掂,当然并没有这样干);另外针对收入这一块,运营人员可能有很多维度需要查询,而且需求是变动的,这个时候数据还是尽量明细保留下来的好,避免需求变动带来的频繁修改代码。
以上,如有不足欢迎留言讨论
错误说明
错误一、scala版本错误
Exception in thread “main” java.lang.NoSuchMethodError: scala.collection.immutable.HashSet$.empty()Lscala/collection/immutable/HashSet;
解决方法:改为使用2.10版本
错误二、客户端调试ssc程序访问kafka异常
WARN - [mytopic_local_ZA6600-1456571483267-f287bec2-leader-finder-thread], Failed to find leader for Set([mytopic,0])
kafka.common.KafkaException: fetching topic metadata for topics [Set(mytopic)] from broker [ArrayBuffer(id:0,host:bj-230,port:9092)] failed
at kafka.client.ClientUtils
.fetchTopicMetadata(ClientUtils.scala:72)atkafka.client.ClientUtils
.fetchTopicMetadata(ClientUtils.scala:93)
at kafka.consumer.ConsumerFetcherManager
LeaderFinderThread.doWork(ConsumerFetcherManager.scala:66)atkafka.utils.ShutdownableThread.run(ShutdownableThread.scala:60)Causedby:java.nio.channels.ClosedChannelExceptionatkafka.network.BlockingChannel.send(BlockingChannel.scala:100)atkafka.producer.SyncProducer.liftedTree1
1(SyncProducer.scala:73)
at kafka.producer.SyncProducer.kafka
producer
SyncProducer$$doSend(SyncProducer.scala:72)
at kafka.producer.SyncProducer.send(SyncProducer.scala:113)
at kafka.client.ClientUtils$.fetchTopicMetadata(ClientUtils.scala:58)
… 3 more
解决方案
仔细看错误返回 from broker [ArrayBuffer(id:0,host:bj-230,port:9092)] failed 这个访问的broker的host竟然是bj-230,是kafka服务所在的服务器,我客户端压根没有这个hostname的映射
这里把它直接修改为ip地址,打开server.properties,增加配置项advertised.host.name=10.10.10.230,重启kafka服务;另一种处理方法也可以在hosts中配置ip映射