Trident中的DRPC实现

一：介绍

1.说明

　　Storm里面引入DRPC主要是利用storm的实时计算能力来并行化CPU intensive的计算。DRPC的storm topology以函数的参数流作为输入，而把这些函数调用的返回值作为topology的输出流。

2.工作机制

　　Distributed RPC是由一个”DPRC Server”协调的(storm自带了一个实现)DRPC服务器协调

　　1) 接收一个RPC请求。

　　2) 发送请求到storm topology

　　3) 从storm topology接收结果。

　　4) 把结果发回给等待的客户端。从客户端的角度来看一个DRPC调用跟一个普通的RPC调用没有任何区别。

3.工作流程

　　

　　客户端给DRPC服务器发送要执行的方法的名字，以及这个方法的参数。

　　实现了这个函数的topology使用 DRPCSpout 从DRPC服务器接收函数调用流。

　　每个函数调用被DRPC服务器标记了一个唯一的id。 这个topology然后计算结果，在topology的最后一个叫做 ReturnResults 的bolt会连接到DRPC服务器，并且把这个调用的结果发送给DRPC服务器(通过那个唯一的id标识)。DRPC服务器用那个唯一id来跟等待的客户端匹配上，唤醒这个客户端并且把结果发送给它。

二：本地DRPC

1.主驱动类

 package com.jun.tridentWithHbase;

 import backtype.storm.Config;
 import backtype.storm.LocalCluster;
 import backtype.storm.LocalDRPC;
 import backtype.storm.StormSubmitter;
 import backtype.storm.generated.AlreadyAliveException;
 import backtype.storm.generated.InvalidTopologyException;
 import backtype.storm.tuple.Fields;
 import backtype.storm.tuple.Values;
 import org.apache.storm.hbase.trident.state.HBaseMapState;
 import storm.trident.Stream;
 import storm.trident.TridentState;
 import storm.trident.TridentTopology;
 import storm.trident.operation.builtin.Count;
 import storm.trident.operation.builtin.MapGet;
 import storm.trident.operation.builtin.Sum;
 import storm.trident.state.OpaqueValue;
 import storm.trident.state.StateFactory;
 import storm.trident.testing.FixedBatchSpout;
 import storm.trident.testing.MemoryMapState;

 public class TridentDemo {
     public static void main(String[] args) throws AlreadyAliveException, InvalidTopologyException {
         TridentTopology tridentTopology=new TridentTopology();
         //模拟数据
         Fields field=new Fields("log","flag");
         FixedBatchSpout spout=new FixedBatchSpout(field,5,
             new Values("168.214.187.214 - - [1481953616092] \"GET /view.php HTTP/1.1\" 200 0 \"http://cn.bing.com/search?q=spark mllib\" \"Mozilla/5.0 (Windows NT 6.1; rv:2.0.1) Gecko/20100101 Firefox/4.0.1\" \"-\"","A"),
             new Values("168.187.202.202 - - [1481953537038] \"GET /IBEIfeng.gif?order_id=1063&orderTime=1481953537038&memberId=4000012340500607&productInfos=10005-2099.48-B-1|10004-1886.62-A-2|10001-961.99-A-1&orderAmt=6834.70 HTTP/1.1\" 200 0 \"-\" \"Mozilla/5.0 (compatible; MSIE 10.0; Windows NT 6.2;Tident/6.0)\" \"-\"","A"),
             new Values("61.30.167.187 - - [1481953539039] \"GET /IBEIfeng.gif?order_id=1064&orderTime=1481953539039&memberId=4000930409959999&productInfos=10007-3329.13-B-1|10009-2607.71-B-1|10002-390.62-A-1|10006-411.00-B-2&orderAmt=7149.46 HTTP/1.1\" 200 0 \"-\" \"Mozilla/5.0 (Linux; Android 4.2.1; Galaxy Nexus Build/JOP40D) AppleWebKit/535.19 (KHTML, like Gecko) Chrome/18.0.1025.166 Mobile Safari/535.19\" \"-\"","A"),
             new Values("30.29.132.190 - - [1481953544042] \"GET /IBEIfeng.gif?order_id=1065&orderTime=1481953544043&memberId=1234568970080798&productInfos=10005-2099.48-B-1|10001-3242.40-C-2|10006-411.00-B-1&orderAmt=8995.28 HTTP/1.1\" 200 0 \"-\" \"Mozilla/5.0 (iPhone; CPU iPhone OS 7_)_3 like Mac OS X) AppleWebKit/537.51.1 (KHTML, like Gecko) Version/7.0 Mobile/11B511 Safari/9537.53\" \"-\"","B"),
             new Values("222.190.187.201 - - [1481953578068] \"GET /IBEIfeng.gif?order_id=1066&orderTime=1481953578068&memberId=3488586887970809&productInfos=10005-2099.48-B-1|10001-2774.16-C-2&orderAmt=7647.80 HTTP/1.1\" 200 0 \"-\" \"Mozilla/5.0 (Windows NT 6.1; rv:2.0.1) Gecko/20100101 Firefox/4.0.1\" \"-\"","B"),
             new Values("72.202.43.53 - - [1481953579069] \"GET /IBEIfeng.gif?order_id=1067&orderTime=1481953579069&memberId=2084859896989877&productInfos=10007-3329.13-B-1|10001-961.99-A-2&orderAmt=5253.10 HTTP/1.1\" 200 0 \"-\" \"Mozilla/5.0 (Linux; Android 4.2.1; Galaxy Nexus Build/JOP40D) AppleWebKit/535.19 (KHTML, like Gecko) Chrome/18.0.1025.166 Mobile Safari/535.19\" \"-\"","B")
         );
         //多次循环
         spout.setCycle(true);
         //流处理
         Stream stream=tridentTopology.newStream("orderAnalyse",spout)
                 //过滤
             .each(new Fields("log"),new ValidLogFilter())
                 //解析
             .each(new Fields("log"), new LogParserFunction(),new Fields("orderId","orderTime","orderAmtStr","memberId"))
                 //投影
             .project(new Fields("orderId","orderTime","orderAmtStr","memberId"))
                 //时间解析
             .each(new Fields("orderTime"),new DateTransFormerFunction(),new Fields("day","hour","minter"))
          ;
         //分流
         //1.基于minter统计订单数量，分组统计
         TridentState state=stream.groupBy(new Fields("minter"))
                 //全局聚合，使用内存存储状态信息
                 .persistentAggregate(new MemoryMapState.Factory(),new Count(),new Fields("orderNumByMinter"));
 //        state.newValuesStream().each(new Fields("minter","orderNumByMinter"),new PrintFilter());

         //2.另一个流,基于分钟的订单金额，局部聚合
         Stream partitionStream=stream.each(new Fields("orderAmtStr"),new TransforAmtToDoubleFunction(),new Fields("orderAmt"))
             .groupBy(new Fields("minter"))
                     //局部聚合
                 .chainedAgg()    //聚合链
             .partitionAggregate(new Fields("orderAmt"),new LocalSum(),new Fields("orderAmtSumOfLocal"))
                 .chainEnd();      //聚合链

         //做一次全局聚合
         TridentState partitionState=partitionStream.groupBy(new Fields("minter"))
                 //全局聚合
                 .persistentAggregate(new MemoryMapState.Factory(),new Fields("orderAmtSumOfLocal"),new Sum(),new Fields("totalOrderAmt"));
         partitionState.newValuesStream().each(new Fields("minter","totalOrderAmt"),new PrintFilter());

         //提交
         Config config=new Config();
         if(args==null || args.length<=0){
             //应该是构建一个DRPC的服务器
             LocalDRPC localDRPC=new LocalDRPC();
             tridentTopology.newDRPCStream("orderDataServer",localDRPC)
                     //参数处理
                 .each(new Fields("args"),new RequestParamsParserFunction(),new Fields("date"))
                     //查询，重要的参数是上面的partitionState
                 .stateQuery(partitionState,new Fields("date"),new MapGet(),new Fields("totalAmtByMinter"))
                     //投影
                 .project(new Fields("date","totalAmtByMinter"));
             //提交任务
             LocalCluster localCluster=new LocalCluster();
             localCluster.submitTopology("tridentDemo",config,tridentTopology.build());
             //获取值
             String jsonResult=localDRPC.execute("orderDataServer","201612171345 201612171345");
             System.out.println("***"+jsonResult+"***");

         }else {
             config.setNumWorkers(2);
             StormSubmitter.submitTopology(args[0],config,tridentTopology.build());
         }
     }
 }

2.请求参数处理类

 package com.jun.tridentWithHbase;

 import backtype.storm.tuple.Values;
 import storm.trident.operation.Function;
 import storm.trident.operation.TridentCollector;
 import storm.trident.operation.TridentOperationContext;
 import storm.trident.tuple.TridentTuple;

 import java.util.Map;

 public class RequestParamsParserFunction implements Function {
     @Override
     public void execute(TridentTuple tridentTuple, TridentCollector tridentCollector) {
         String parameters=tridentTuple.getStringByField("args");
         String[] params=parameters.split(" ");
         for (String param:params){
             tridentCollector.emit(new Values(param));
         }
     }

     @Override
     public void prepare(Map map, TridentOperationContext tridentOperationContext) {

     }

     @Override
     public void cleanup() {

     }
 }

3.效果

　　

三：集群模式的DRPC　

1.主驱动类

  config.setNumWorkers(2);
             //集群上构建DRPC服务器
             tridentTopology.newDRPCStream("orderDataServer")
                     //参数处理
                     .each(new Fields("args"),new RequestParamsParserFunction(),new Fields("date"))
                     //查询，重要的参数是上面的partitionState
                     .stateQuery(partitionState,new Fields("date"),new MapGet(),new Fields("totalAmtByMinter"))
                     //投影
                     .project(new Fields("date","totalAmtByMinter"));
             StormSubmitter.submitTopology(args[0],config,tridentTopology.build());

2.配置DRPC服务和端口

　　

3.启动storm

4.启动Drpc进程

　　在drpc.servers参数所指定的服务器上。

　　命令：nohup bin/storm drpc >>/dev/null 2>&1 &

　　查看端口是否正常打开：netstat -tlnup | grep 3772

5.将jar包提交到集群上

6.编写客户端

 package com.jun.tridentWithKafka;

 import backtype.storm.generated.DRPCExecutionException;
 import backtype.storm.utils.DRPCClient;
 import org.apache.thrift7.TException;

 public class DrpcClientDemo {
     public static void main(String[] args) {
         DRPCClient drpcClient=new DRPCClient("linux-hadoop01.ibeifeng.com",3772);
         try {
             String jsonResult=drpcClient.execute("orderDataServer","201612171345 201612171345");
             System.out.println("==="+jsonResult+"===");
         } catch (TException e) {
             e.printStackTrace();
         } catch (DRPCExecutionException e) {
             e.printStackTrace();
         }
     }
 }