flume+hbase+kafka集成部署

目录

一、架构图

二、flume的安装和配置

1. 下载安装

2. 配置

2.1 日志收集节点配置

2.2 日志聚集节点配置

2.3 Flume的二次开发

三、模拟程序开发

四、服务启动测试

1. 服务启动

1.1 zookeeper

1.2 HDFS

1.3 Hbase

1.4 kafka

2. 测试

---

Flume是一个高可用的，高可靠的，分布式的海量日志采集、聚合和传输的系统，Flume支持在日志系统中定制各类数据发送方，用于收集数据；同时，Flume提供对数据进行简单处理，并写到各种数据接受方。flume包含三个部分：source、channel和sink。

本实例将使用flume收集日志数据分别写入到kafka(topic：logs)和hbase(table：logs)中，日志来源于：搜狗实验室，我们将用程序模拟webserver来写日志(webserverlogs文件)。

安装配置kafka集群请参阅：kafka的配置和分布式部署，安装配置Hbase数据库请参阅：Hbase的配置和分布式部署

说明：三台机器的主机名分别为：bigdata.centos01、bigdata.centos02、bigdata.centos03

一、架构图

说明：centos02和centos03作为flume日志收集节点，centos01将两台机器收集到的日志进行聚合，然后再分别写入Kafka和Hbase数据库中。

二、flume的安装和配置

1. 下载安装

wget http://archive.apache.org/dist/flume/1.7.0/apache-flume-1.7.0-bin.tar.gz

2. 配置

2.1 日志收集节点配置

如上图所示，修改centos02和centos03上flume节点配置

• conf/flume-env.sh修改

export JAVA_HOME=/opt/modules/jdk8

• conf/flume-conf.properties修改

a1.sources = r1
a1.channels = c1
a1.sinks = k1

# source
# exec source：运行unix命令，实时读取webserverlogs日志文件，加入到flume
a1.sources.r1.type = exec
a1.sources.r1.command = tail -F /opt/datas/webserverlogs
a1.sources.r1.channels = c1

# channel
a1.channels.c1.type = memory

# sink
a1.sinks.k1.type = avro
a1.sinks.k1.channel = c1
a1.sinks.k1.hostname = bigdata.centos01
a1.sinks.k1.port = 5555

2.2 日志聚集节点配置

如上图所示，修改centos01上flume节点配置

• conf/flume-env.sh修改

export JAVA_HOME=/opt/modules/jdk8
export HADOOP_HOME=/opt/modules/hadoop-2.5.0
export HBASE_HOME=/opt/modules/hbase-0.98.6-cdh5.3.9

• conf/flume-conf.properties修改

agent.sources = avroSource
agent.channels = kafkaC hbaseC
agent.sinks = kafkaSink hbaseSink

#********************* flume + hbase 集成*********************
# source
agent.sources.avroSource.type = avro
agent.sources.avroSource.channels = hbaseC kafkaC
agent.sources.avroSource.bind = bigdata.centos01
agent.sources.avroSource.port = 5555

# channel
agent.channels.hbaseC.type = memory
agent.channels.hbaseC.capacity = 100000
agent.channels.hbaseC.transactionCapacity = 100000
agent.channels.hbaseC.keep-alive = 10

# sink
agent.sinks.hbaseSink.type = asynchbase
agent.sinks.hbaseSink.table = logs
agent.sinks.hbaseSink.columnFamily = info
agent.sinks.hbaseSink.serializer = org.apache.flume.sink.hbase.SimpleAsyncHbaseEventSerializer
agent.sinks.hbaseSink.channel = hbaseC
# 自定义列
agent.sinks.hbaseSink.serializer.payloadColumn=time,userid,searchname,retorder,cliorder,url


#********************* flume + kafka 集成*********************
# channel
agent.channels.kafkaC.type = memory
agent.channels.kafkaC.capacity = 100000
agent.channels.kafkaC.transactionCapacity = 100000
agent.channels.kafkaC.keep-alive = 10

# sink
agent.sinks.kafkaSink.channel = kafkaC
agent.sinks.kafkaSink.type = org.apache.flume.sink.kafka.KafkaSink
agent.sinks.kafkaSink.kafka.topic = logs
agent.sinks.kafkaSink.kafka.bootstrap.servers = bigdata.centos01:9092,bigdata.centos02:9092,bigdata.centos03:9092
agent.sinks.kafkaSink.kafka.producer.acks = 1
agent.sinks.kafkaSink.kafka.flumeBatchSize = 50

2.3 Flume的二次开发

• 下载源码

http://archive.apache.org/dist/flume/1.7.0/apache-flume-1.7.0-src.tar.gz

• 代码修改

由于上面centos01 flume hbasesink配置的SimpleAsyncHbaseEventSerializer类不能满足我们的需求，它默认是往Hbase里面写一列，我们的需求是分隔日志数据分别存储到我们定义的不同列，故需要对其进行二次开发。主要修改了getActions函数。

package org.apache.flume.sink.hbase;

import com.google.common.base.Charsets;

import org.apache.flume.Context;
import org.apache.flume.Event;
import org.apache.flume.FlumeException;
import org.apache.flume.conf.ComponentConfiguration;
import org.apache.flume.sink.hbase.SimpleHbaseEventSerializer.KeyType;
import org.hbase.async.AtomicIncrementRequest;
import org.hbase.async.PutRequest;

import java.util.ArrayList;
import java.util.List;

public class SimpleAsyncHbaseEventSerializer implements AsyncHbaseEventSerializer {
    private byte[] table;
    private byte[] cf;
    private byte[] payload;
    private byte[] payloadColumn;
    private byte[] incrementColumn;
    private String rowPrefix;
    private byte[] incrementRow;
    private KeyType keyType;

    @Override
    public void initialize(byte[] table, byte[] cf) {
        this.table = table;
        this.cf = cf;
    }

    @Override
    public List getActions() {
        List actions = new ArrayList();
        if (payloadColumn != null) {
            try {
                String[] columns = new String(payloadColumn).split(",");
                String[] values = new String(payload).split(",");
                if (columns.length != values.length) {
                    return actions;
                }

                String rawKey = values[0] + values[1] + System.currentTimeMillis();
                for (int i = 0; i < columns.length; i++) {
                    String column = columns[i];
                    String value = values[i];

                    PutRequest putRequest = new PutRequest(table, rawKey.getBytes("utf-8"), cf,
                            column.getBytes(), value.getBytes("utf-8"));
                    actions.add(putRequest);
                }

            } catch (Exception e) {
                throw new FlumeException("Could not get row key!", e);
            }
        }
        return actions;
    }

    public List getIncrements() {
        List actions = new ArrayList();
        if (incrementColumn != null) {
            AtomicIncrementRequest inc = new AtomicIncrementRequest(table,
                    incrementRow, cf, incrementColumn);
            actions.add(inc);
        }
        return actions;
    }

    @Override
    public void cleanUp() {
        // TODO Auto-generated method stub
    }

    @Override
    public void configure(Context context) {
        String pCol = context.getString("payloadColumn", "pCol");
        String iCol = context.getString("incrementColumn", "iCol");
        rowPrefix = context.getString("rowPrefix", "default");
        String suffix = context.getString("suffix", "uuid");
        if (pCol != null && !pCol.isEmpty()) {
            if (suffix.equals("timestamp")) {
                keyType = KeyType.TS;
            } else if (suffix.equals("random")) {
                keyType = KeyType.RANDOM;
            } else if (suffix.equals("nano")) {
                keyType = KeyType.TSNANO;
            } else {
                keyType = KeyType.UUID;
            }
            payloadColumn = pCol.getBytes(Charsets.UTF_8);
        }
        if (iCol != null && !iCol.isEmpty()) {
            incrementColumn = iCol.getBytes(Charsets.UTF_8);
        }
        incrementRow = context.getString("incrementRow", "incRow").getBytes(Charsets.UTF_8);
    }

    @Override
    public void setEvent(Event event) {
        this.payload = event.getBody();
    }

    @Override
    public void configure(ComponentConfiguration conf) {
        // TODO Auto-generated method stub
    }

}

将lib目录的flume-ng-hbase-sink-1.7.0.jar删除，把flume-ng-hbase-sink重新打包重命名为flume-ng-hbase-sink-1.7.0.jar，上传至flume的lib目录。我打包的jar包：flume-ng-hbase-sink.jar

三、模拟程序开发

通过模拟程序将搜狗实验室的日志文件逐行的写入到webserverlogs文件，供日志收集的flume收集。

package com.wmh.writeread;

import java.io.*;

public class ReadWrite {

    public static void main(String[] args) {
        String readFileName = args[0];
        String writeFileName = args[1];

        try{
            System.out.println("执行中...");
            process(readFileName,writeFileName);
            System.out.println("执行完成！");
        }catch (Exception e){
            System.out.println("执行失败！");
        }
    }

    private static void process(String readFileName, String writeFileName) {
        File readFile = new File(readFileName);
        if (!readFile.exists()){
            System.out.println(readFileName+"不存在，请检查路径！");
            System.exit(1);
        }
        File writeFile = new File(writeFileName);

        BufferedReader br = null;
        FileInputStream fis = null;
        FileOutputStream fos = null;
        long count = 1l;
        try{
            fis = new FileInputStream(readFile);
            fos = new FileOutputStream(writeFile);
            br = new BufferedReader(new InputStreamReader(fis));
            String line = "";
            while ((line = br.readLine()) != null){
                fos.write((line + "\n").getBytes("utf-8"));
                fos.flush();
                Thread.sleep(100);
                System.out.println(String.format("row:[%d]>>>>>>>>>> %s",count++,line));
            }

        }catch (Exception e){
            try {
                fis.close();
                fos.close();
                br.close();
            } catch (IOException ex) {
                System.out.println("关闭资源异常！");
                System.exit(1);
            }
        }

    }

}

四、服务启动测试

1. 服务启动

1.1 zookeeper

启动centos01、centos02、centos03的zookeeper

bin/zkServer.sh start

1.2 HDFS

启动centos01的namenode、datanode，centos02和centos03的datanode

sbin/start-dfs.sh

1.3 Hbase

• 服务启动

启动centos01的master、regionserver，centos02和centos03的regionserver

bin/start-hbase.sh

• 创建表

# 进入命令行操作
bin/hbase shell

# 创建表
> create "logs","info"

1.4 kafka

• 服务启动

启动centos01、centos02、centos03的kafka

bin/kafka-server-start.sh config/server.properties

• 创建topic

创建flume配置的名为logs的topic

# --replication-factor 副本数，为了每个kafka可以单独对外服务，该项配置值为集群机器数
bin/kafka-topics.sh --create --zookeeper bigdata.centos01:2181,bigdata.centos02:2181,bigdata.centos03:2181 --replication-factor 3 --partitions 1 --topic logs

• 启动consumer

启动任意一台机器上面的kafka consumer，供测试使用。

bin/kafka-console-consumer.sh --zookeeper bigdata.centos01:2181,bigdata.centos02:2181,bigdata.centos03:2181 --topic logs --from-beginning

1.5 flume

• 启动centos01日志聚集flume

bin/flume-ng agent --conf conf --conf-file conf/flume-conf.properties -n agent  -Dflume.root.logger=INFO,console

• 启动centos01日志收集flume

bin/flume-ng agent --conf conf --conf-file conf/flume-conf.properties -n a1  -Dflume.root.logger=INFO,console

主要区别在于：agent命名不一样，centos01命名为agent，centos02和centos03的命名为a1

1.6 模拟程序启动

将上面的模拟程序类打包成jar

#formatedSougouLog是经格式化的数据，将\t和空格替换成逗号
java -jar ReadWrite.jar formatedSougouLog webserverlogs

2. 测试

• hbase插入的数据

• kafka获取的数据