Flume客户端flume-ng-log4jappender 负载平衡（LoadBalancingLog4jAppender）

主机DNS配置：

192.168.177.167 machine-1
192.168.177.168 machine-2
192.168.177.158 machine-0
192.168.177.174 hadoop-master hbase-master

hadoop-maser 和machine-2当主机，其它机器当做collector机，存储在HDFS中。

hadoop-master和machine-2机上的flume配置：

agent.sources=s1
agent.channels=c1
agent.sinks=k1 k2

agent.sinkgroups = g1 
agent.sinkgroups.g1.sinks = k1 k2  
agent.sinkgroups.g1.processor.type = load_balance  
agent.sinkgroups.g1.processor.selector = round_robin  
agent.sinkgroups.g1.processor.backoff = true  


agent.sources.s1.type=avro
agent.sources.s1.channels=c1
agent.sources.s1.bind=0.0.0.0
agent.sources.s1.port=51515
agent.sources.s1.interceptors=i1
agent.sources.s1.interceptors.i1.type=timestamp


agent.channels.c1.type=jdbc


agent.sinks.k1.channel = c1  
agent.sinks.k1.type = avro  
agent.sinks.k1.hostname = machine-0
agent.sinks.k1.port = 51515
agent.sinks.k2.channel = c1  
agent.sinks.k2.type = avro
agent.sinks.k2.hostname = machine-1
agent.sinks.k2.port = 51515 

machine-1 和machine-0的flume配置：

agent.sources=s1
agent.channels=c1
agent.sinks=k1


agent.sources.s1.type=avro
agent.sources.s1.channels=c1
agent.sources.s1.bind=0.0.0.0
agent.sources.s1.port=51515


agent.channels.c1.type=jdbc


agent.sinks.k1.type=hdfs
agent.sinks.k1.channel=c1
agent.sinks.k1.hdfs.path=/flume/%Y/%m
agent.sinks.k1.hdfs.filePrefix=flume
agent.sinks.k1.hdfs.fileSuffix=.log
agent.sinks.k1.hdfs.rollInterval=3600
agent.sinks.k1.hdfs.rollCount=0
agent.sinks.k1.hdfs.rollSize=0
agent.sinks.k1.hdfs.fileType=DataStream
agent.sinks.k1.hdfs.writeFormat=Text
agent.sinks.k1.hdfs.useLocalTimeStamp=false

log4j的配置：

# File Appender rootLog
log4j.rootLogger=DEBUG,stdout,rootLog


#console configure for DEV environment
log4j.appender.stdout=org.apache.log4j.ConsoleAppender
log4j.appender.stdout.layout=org.apache.log4j.PatternLayout
log4j.appender.stdout.layout.ConversionPattern=%d{yyyy-MM-dd HH:mm:ss,SSS} %-5p (%c:%L) - %m%n


log4j.appender.rootLog=org.apache.log4j.RollingFileAppender
log4j.appender.rootLog.File= rootLog.log
log4j.appender.rootLog.MaxFileSize=5000KB
log4j.appender.rootLog.MaxBackupIndex=20
log4j.appender.rootLog.layout=org.apache.log4j.PatternLayout
log4j.appender.rootLog.layout.ConversionPattern=%d{yyyy-MM-dd HH:mm:ss,SSS} %-5p (%c:%L) - %m%n


# File Appender boentel
#log4j.logger.com.boentel=DEBUG,boentel
#log4j.additivity.com.boentel=true
#log4j.appender.boentel=org.apache.log4j.RollingFileAppender
#log4j.appender.boentel.File= boentel.log
#log4j.appender.boentel.MaxFileSize=2000KB
#log4j.appender.boentel.MaxBackupIndex=20
#log4j.appender.boentel.layout=org.apache.log4j.PatternLayout
#log4j.appender.boentel.layout.ConversionPattern=%d{yyyy-MM-dd HH:mm:ss,SSS} %-5p (%c:%L) - %m%n


log4j.logger.com.loadbalance= DEBUG,loadbalance
log4j.additivity.com.loadbalance= true


log4j.appender.loadbalance = org.apache.flume.clients.log4jappender.LoadBalancingLog4jAppender
log4j.appender.loadbalance.Hosts =machine-2:51515 hadoop-master:51515
#log4j.appender.loadbalance.UnsafeMode = true
log4j.appender.out2.MaxBackoff = 30000
 #FQDN RANDOM ,default is ROUND_ROBIN
log4j.appender.loadbalance.Selector = RANDOM
log4j.appender.loadbalance.layout=org.apache.log4j.PatternLayout
log4j.appender.loadbalance.layout.ConversionPattern=%d{yyyy-MM-dd HH:mm:ss,SSS} %-5p (%c:%L) - %m%n

测试代码：

import java.util.Date;
import java.util.concurrent.Executors;
import java.util.concurrent.ScheduledExecutorService;
import java.util.concurrent.TimeUnit;


import org.apache.log4j.Logger;






public class Worker implements Runnable{


	
	private static final Logger LOG = Logger.getLogger(Worker.class);  
	private String command;
	
	/**
	 * @param args
	 */
	public static void main(String[] args) {
        new Worker("0").init();
	}


	public void init(){
		int numWorkers = 1;
        int threadPoolSize = 3 ;


        ScheduledExecutorService scheduledThreadPool = Executors.newScheduledThreadPool(threadPoolSize);
        
        //schedule to run after sometime
        System.out.println("Current Time = "+new Date());
        Worker worker = null;
        for(int i=0; i< numWorkers; i++){
            try {
				Thread.sleep(1000);
			} catch (InterruptedException e) {
				e.printStackTrace();
			}
            worker = new Worker("do heavy processing");
//	            scheduledThreadPool.schedule(worker, 10, TimeUnit.SECONDS);
            //scheduleAtFixedRate
//	            scheduledThreadPool.scheduleAtFixedRate(worker, 0, 1, TimeUnit.SECONDS);
            scheduledThreadPool.scheduleWithFixedDelay(worker, 5, 10,
                    TimeUnit.SECONDS);
        
        }
         
        //add some delay to let some threads spawn by scheduler
        try {
			Thread.sleep(30000);
		} catch (InterruptedException e) {
			e.printStackTrace();
		}
         
        scheduledThreadPool.shutdown();
        while(!scheduledThreadPool.isTerminated()){
            //wait for all tasks to finish
        }
        LOG.info("Finished all threads");
	}
	public Worker(String command){
		this.command = command;
	}
	@Override
    public void run() {
        LOG.info(Thread.currentThread().getName()+" Start. Command = "+command);
        processCommand();
        LOG.info(Thread.currentThread().getName()+" End.");
    }
 
    private void processCommand() {
        try {
            for(int i = 1000; i < 1200; i++){
            	LOG.info("sequence:" + i);
            }
            Thread.sleep(5000);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
    }
    
    @Override
    public String toString(){
        return this.command;
    }
}

小结：

最终能实现负载均衡的作用，但是，性能上还有些欠缺。

当一台机死掉时，客户端将尝试不断链接，影响到数据传送到其它机子上。当死掉的机器恢复后，客户端备份的数据会重新发送到flume agent。数据正确性是达到了，但是，万一这个app当掉了，对应的日志信息不就丢了吗？这是一个问题，有待进一步的改进。