Hadoop日志文件分析系统

Hadoop日志分析系统

项目需求：

需要统计一下线上日志中某些信息每天出现的频率，举个简单的例子，统计线上每天的请求总数和异常请求数。线上大概几十台

服务器，每台服务器大概每天产生4到5G左右的日志，假设有30台，每台5G的，一天产生的日志总量为150G。

处理方案：

方案1：传统的处理方式，写个JAVA日志分析代码，部署到每台服务器进行处理，这种方式部署起来耗时费力，又不好维护。

方案2：采用Hadoop分布式处理，日志分析是Hadoop集群系统的拿手好戏。150G每天的日志也算是比较大的数据量了，搭个简

单的Hadoop集群来处理这些日志是再好不过的了。

Hadoop集群的搭建：

参见这两篇文章：http://www.cnblogs.com/cstar/archive/2012/12/16/2820209.html

http://www.cnblogs.com/cstar/archive/2012/12/16/2820220.html

我们这里的集群就采用了两台机器，配置每台8核，32G内存，500G磁盘空间。

日志准备工作：

由于日志分散在各个服务器，所以我们先需要将所有的日志拷贝到我们的集群系统当中，这个可以通过linux服务器下rsync或者scp

服务来执行。这里我们通过scp服务来拷贝，由于都是内网的机器，所以拷贝几个G的日志可以很快就完成。下面是拷贝日志的脚本，脚本

还是有一些需要注意的地方，我们只需要拷贝前一天的数据，实际保存的数据可能是好几天的，所以我们只要把我们需要的这一天的数据

SCP过去就可以了。

#!/bin/sh
workdir=/home/myproj/bin/log/
files=`ls $workdir`
pre1date=`date  +"%Y%m%d" -d  "-1 days"`
pre1date1=`date  +"%Y-%m-%d" -d  "-1 days"`
curdate=`date  +"%Y%m%d"`
hostname=`uname -n`
echo $pre1date $curdate
uploadpath="/home/hadoop/hadoop/mytest/log/"$pre1date1"/"$hostname
echo $uploadpath
cd $workdir
mintime=240000
secondmintime=0
for file in $files;do
  filedate=`stat $file | grep Modify| awk '{print $2}' |sed -e 's/-//g'`
  filetime=`stat $file | grep Modify| awk '{print $3}' |cut -d"." -f1 | sed -e 's/://g'| sed 's/^0\+//'`
  if [ $filedate -eq $curdate ]; then
   if [ $filetime -lt $mintime ]; then
        secondmintime=$mintime
	mintime=$filetime
   fi
  fi
done
echo "mintime:"$mintime
step=1000
mintime=`expr $mintime + $step`
echo "mintime+1000:"$mintime
for file in $files;do
  filedate=`stat $file | grep Modify| awk '{print $2}' |sed -e 's/-//g'`
  filetime=`stat $file | grep Modify| awk '{print $3}' |cut -d"." -f1 | sed -e 's/://g'| sed 's/^0\+//'`
  filename=`echo $file | cut -c 1-8`
  startchars="info.log"
  #echo $filename
  if [ $filename == $startchars ]; then
    if [ $filedate -eq $pre1date ]; then
     scp -rp $file dir@antix2:$uploadpath
     #echo $file
    elif [ $filedate -eq $curdate ]; then
      if [ $filetime -lt $mintime ]; then
        scp -rp $file dir@antix2:$uploadpath
        #echo $file
      fi
    fi
  fi
  #echo $filedate $filetime
done

MapReduce代码

   接下来就是编写MapReduce的代码了。使用Eclipse环境来编写，需要安装hadoop插件，我们hadoop机器采用的是1.1.1版本，所以插

件使用hadoop-eclipse-plugin-1.1.1.jar，将插件拷贝到eclipse的plugins目录下就可以了。然后新建一个MapReduce项目：

工程新建好了然后我们就可以编写我们的MapReduce代码了。

import java.io.IOException;
import java.text.SimpleDateFormat;
import java.util.Date;

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.IntWritable;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.Mapper;
import org.apache.hadoop.mapreduce.Reducer;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
import org.apache.hadoop.util.GenericOptionsParser;

public class LogAnalysis {

    public static class LogMapper
    extends Mapper<LongWritable, Text, Text, IntWritable>{

        private final static IntWritable one = new IntWritable(1);
        private Text word = new Text();
        private Text hourWord = new Text();
        public void map(LongWritable key, Text value, Context context
                ) throws IOException, InterruptedException {
            String line = value.toString();
            SimpleDateFormat formatter2 = new SimpleDateFormat("yy-MM-dd");
            java.util.Date d1 =new Date();
            d1.setTime(System.currentTimeMillis()-1*24*3600*1000);
            String strDate =formatter2.format(d1);
            if(line.contains(strDate)){
                 String[] strArr = line.split(",");
                 int len = strArr[0].length();
                 String time = strArr[0].substring(1,len-1);

                 String[] timeArr = time.split(":");
                 String strHour = timeArr[0];
                 String hour = strHour.substring(strHour.length()-2,strHour.length());
                 String hourKey = "";
                if(line.contains("StartASocket")){
                    word.set("SocketCount");
                    context.write(word, one);
                    hourKey = "SocketCount:" + hour;
                    hourWord.set(hourKey);
                    context.write(hourWord, one);
                    word.clear();
                    hourWord.clear();
                }
                if(line.contains("SocketException")){
                    word.set("SocketExceptionCount");
                    context.write(word, one);
                    hourKey = "SocketExceptionCount:" + hour;
                    hourWord.set(hourKey);
                    context.write(hourWord, one);
                    word.clear();
                    hourWord.clear();
                }
                          
        }
}

public static class LogReducer
extends Reducer<Text,IntWritable,Text,IntWritable> {
    private IntWritable result = new IntWritable();

    public void reduce(Text key, Iterable<IntWritable> values,
            Context context
            ) throws IOException, InterruptedException {
        int sum = 0;
        for (IntWritable val : values) {
            sum += val.get();
        }
        result.set(sum);
        context.write(key, result);
    }
}
public static int run(String[] args) throws Exception{

    Configuration conf = new Configuration();
    String[] otherArgs = new GenericOptionsParser(conf, args).getRemainingArgs();
    if (otherArgs.length != 2) {
        System.err.println("Usage: loganalysis <in> <out>");
        System.exit(2);
    }
    FileSystem fileSys = FileSystem.get(conf);
    String inputPath = "input/" + args[0];
    fileSys.copyFromLocalFile(new Path(args[0]), new Path(inputPath));//将本地文件系统的文件拷贝到HDFS中
    Job job = new Job(conf, "loganalysis");
    job.setJarByClass(LogAnalysis.class);
    job.setMapperClass(LogMapper.class);
    job.setCombinerClass(LogReducer.class);
    job.setReducerClass(LogReducer.class);
    // 设置输出类型
    job.setOutputKeyClass(Text.class);
    job.setOutputValueClass(IntWritable.class);
    FileInputFormat.addInputPath(job, new Path(inputPath));
    FileOutputFormat.setOutputPath(job, new Path(otherArgs[1]));

    Date startTime = new Date();
    System.out.println("Job started: " + startTime);
    int ret = job.waitForCompletion(true)? 0 : 1;
    fileSys.copyToLocalFile(new Path(otherArgs[1]), new Path(otherArgs[1]));
    fileSys.delete(new Path(inputPath), true);
    fileSys.delete(new Path(otherArgs[1]), true);   

    Date end_time = new Date();
    System.out.println("Job ended: " + end_time);
    System.out.println("The job took " + (end_time.getTime() - startTime.getTime()) /1000 + " seconds.");
    return ret;
}

public static void main(String[] args)
{
    try
    {
        int ret = run(args);
        System.exit(ret);
    } catch (Exception e)
    {
        e.printStackTrace();
        System.out.println(e.getMessage());
    }
}
}

部署到Hadoop集群：

代码完成后测试没有问题后，部署到集群当中去执行，我们有几十台服务器，所以每台的服务器的日志当成一个任务来执行。

workdir="/home/hadoop/hadoop/mytest"
cd $workdir
pre1date=`date  +"%Y-%m-%d" -d  "-1 days"`
servers=(mach1 mach2 mach3 )
for i in ${servers[@]};do
  inputPath="log/"$pre1date"/"$i
  outputPath="output/log/"$pre1date"/"$i
  echo $inputPath $outputPath
  echo "start job "$i" date:"`date`
  hadoop jar  LogAnalysis.jar loganalysis $inputPath $outputPath
  echo "end job "$i" date:"`date`
done