1、自定义FlumeKafkaSink

package com.panguoyuan.flume.sink;

import java.util.Iterator;
import java.util.Map;
import java.util.Map.Entry;
import java.util.Properties;

import kafka.javaapi.producer.Producer;
import kafka.producer.KeyedMessage;
import kafka.producer.ProducerConfig;

import org.apache.flume.Channel;
import org.apache.flume.Context;
import org.apache.flume.Event;
import org.apache.flume.EventDeliveryException;
import org.apache.flume.Transaction;
import org.apache.flume.conf.Configurable;
import org.apache.flume.sink.AbstractSink;

 
public class KafkaSink extends AbstractSink implements Configurable {
   
    private String topic;
    private Producer<String, String> producer;
 
    public void configure(Context context) {
        Properties props = new Properties();
        Map<String,String> map = context.getParameters();
        Iterator iter = map.entrySet().iterator();
        String  metadataBrokerList = map.get("metadata.broker.list");
        props.setProperty("metadata.broker.list", metadataBrokerList);
        
        topic = map.get("custom.topic.name");
        if(topic == null || "".equals(topic)){
        	topic = "kafka-defualt-topic-name";
        }
        props.setProperty("serializer.class", "kafka.serializer.StringEncoder");
        props.put("partitioner.class", "com.panguoyuan.flume.sink.HashSimplePartitioner");
        props.put("zookeeper.connect", "ip1:2181,ip2:2181,ip3:2181");
        props.setProperty("num.partitions", "3"); //
        props.put("request.required.acks", "1");
//        props.put("replication-factor", "3");
        
        ProducerConfig config = new ProducerConfig(props);
        producer = new Producer<String, String>(config);
       
        
        while(iter.hasNext()){
        	Entry<String, String> entry = (Entry)iter.next();
        	System.out.println("key="+entry.getKey()+"  value="+entry.getValue());
        }
        
        System.out.println("KafkaSink初始化完成.");
 
    }
 
    public Status process() throws EventDeliveryException {
        Channel channel = getChannel();
        Transaction tx = channel.getTransaction();
        try {
            tx.begin();
            Event e = channel.take();
            if (e == null) {
                tx.rollback();
                return Status.BACKOFF;
            }
            KeyedMessage<String, String> data = new KeyedMessage<String, String>(topic, new String(e.getBody()));
            producer.send(data);
            System.out.println("flume向kafka发送消息：" + new String(e.getBody()));
            tx.commit();
            return Status.READY;
        } catch (Exception e) {
        	System.err.println("Flume KafkaSinkException:"+ e);
            tx.rollback();
            return Status.BACKOFF;
        } finally {
            tx.close();
        }
    }
}

2、自定义Partitioner

import kafka.producer.Partitioner;
import kafka.utils.VerifiableProperties;

/**
 * 自定义Kafka Partitioner
 * 
 * @Author 潘国远
 * 
 * @Time 2015-02-07
 * 
 */
public class HashSimplePartitioner implements Partitioner {
	public HashSimplePartitioner(VerifiableProperties props) {

	}


	@Override
	public int partition(Object obj, int numPartitions) {
		int partition = 0;
		partition = obj.hashCode() % numPartitions;
		return partition;
	}

}

3、编写flume-kafka接受端

#bin/flume-ng agent --conf conf --conf-file conf/flume-accept-conf.properties --name producer -Dflume.root.logger=INFO,console
#agent 
producer.sources = s
producer.channels = c
producer.sinks = r

#source 
producer.sources.s.type = avro
producer.sources.s.bind = 192.168.1.100
producer.sources.s.port = 44444
producer.sources.s.channels = c

producer.sinks.r.type = com.panguoyuan.flume.sink.KafkaSink
producer.sinks.r.metadata.broker.list=ip1:9092,ip2:9092,ip3:9092
producer.sinks.r.partition.key=0
producer.sinks.r.partitioner.class=com.panguoyuan.flume.sink.HashSimplePartitioner
producer.sinks.r.serializer.class=kafka.serializer.StringEncoder
producer.sinks.r.request.required.acks=0
producer.sinks.r.max.message.size=1000000
producer.sinks.r.producer.type=sync
producer.sinks.r.custom.encoding=UTF-8
producer.sinks.r.custom.topic.name=flume-kafka-topic
producer.sinks.r.keep-alive = 10
producer.sinks.r.channel = c

producer.channels.c.type = memory
producer.channels.c.capacity = 10000
producer.channels.c.transactionCapacity = 10000

4、编写flume数据发送端口

send.sources = source1 source2
send.channels = c
send.sinks = s

send.sources.source1.type = exec
send.sources.source1.command = tail -F /app/log1.log
send.sources.source1.channels = c 

send.sources.source2.type = exec
send.sources.source2.command = tail -F /app/log2.log
send.sources.source2.channels = c 

send.channels.c.type = memory
send.channels.c.capacity = 10000
send.channels.c.transactionCapacity = 10000 


send.sinks.s.type = avro
send.sinks.s.hostname = 192.168.1.100
send.sinks.s.port = 44444
send.sinks.s.channel = c
send.sinks.s.keep-alive = 10