1 使用Hive或者自定义MR实现如下逻辑
product_no lac_id moment start_time user_id county_id staytime city_id
13429100031 22554 8 2013-03-11 08:55:19.151754088 571 571 282 571
13429100082 22540 8 2013-03-11 08:58:20.152622488 571 571 270 571
13429100082 22691 8 2013-03-11 08:56:37.149593624 571 571 103 571
13429100087 22705 8 2013-03-11 08:56:51.139539816 571 571 220 571
13429100087 22540 8 2013-03-11 08:55:45.150276800 571 571 66 571
13429100082 22540 8 2013-03-11 08:55:38.140225200 571 571 133 571
13429100140 26642 9 2013-03-11 09:02:19.151754088 571 571 18 571
13429100082 22691 8 2013-03-11 08:57:32.151754088 571 571 287 571
13429100189 22558 8 2013-03-11 08:56:24.139539816 571 571 48 571
13429100349 22503 8 2013-03-11 08:54:30.152622440 571 571 211 571
字段解释:
product_no:用户手机号;
lac_id:用户所在基站;
start_time:用户在此基站的开始时间;
staytime:用户在此基站的逗留时间。
需求描述:
根据lac_id和start_time知道用户当时的位置,根据staytime知道用户各个基站的逗留时长。根据轨迹合并连续基站的staytime。
最终得到每一个用户按时间排序在每一个基站驻留时长
期望输出举例:
13429100082 22540 8 2013-03-11 08:58:20.152622488 571 571 270 571
13429100082 22691 8 2013-03-11 08:56:37.149593624 571 571 390 571
13429100082 22540 8 2013-03-11 08:55:38.140225200 571 571 133 571
13429100087 22705 8 2013-03-11 08:56:51.139539816 571 571 220 571
13429100087 22540 8 2013-03-11 08:55:45.150276800 571 571 66 571
2 Linux脚本能力考察
2.1 请随意使用各种类型的脚本语言实现:批量将指定目录下的所有文件中的$HADOOP_HOME$替换成/home/ocetl/app/hadoop
2.2 假设有10台主机,H1到H10,在开启SSH互信的情况下,编写一个或多个脚本实现在所有的远程主机上执行脚本的功能
例如:runRemoteCmd.sh "ls -l"
期望结果:
H1:
XXXXXXXX
XXXXXXXX
XXXXXXXX
H2:
XXXXXXXX
XXXXXXXX
XXXXXXXX
H3:
...
3 Hadoop基础知识与问题分析的能力
3.1 描述一下hadoop中,有哪些地方使用了缓存机制,作用分别是什么
3.2 请描述https://issues.apache.org/jira/browse/HDFS-2379说的是什么问题,最终解决的思路是什么?
4 MapReduce开发能力
请参照wordcount实现一个自己的map reduce,需求为:
a 输入文件格式:
xxx,xxx,xxx,xxx,xxx,xxx,xxx
b 输出文件格式:
xxx,20
xxx,30
xxx.40
c 功能:根据命令行参数统计输入文件中指定关键字出现的次数,并展示出来
例如:hadoop jar xxxxx.jar keywordcount xxx,xxx,xxx,xxx(四个关键字)
5 MapReduce优化
请根据第五题中的程序, 提出如何优化MR程序运行速度的思路
6 Linux操作系统知识考察
请列举曾经修改过的/etc下的配置文件,并说明修改要解决的问题?
7 Java开发能力
7.1 写代码实现1G大小的文本文件,行分隔符为\x01\x02,统计一下该文件中的总行数,要求注意边界情况的处理
7.2 请描述一下在开发中如何对上面的程序进行性能分析,对性能进行优化的过程
答案如下:
1. 考虑后,决定使用 MR 来实现,于是使用Java,用一个MR Job完成这个事情:
- package org.aboutyun;
- import org.apache.commons.lang.StringUtils;
- import org.apache.hadoop.conf.Configuration;
- import org.apache.hadoop.fs.Path;
- 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.input.TextInputFormat;
- import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
- import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;
- import java.io.IOException;
- import java.text.ParseException;
- import java.text.SimpleDateFormat;
- import java.util.ArrayList;
- import java.util.Collections;
- import java.util.Comparator;
- public class TimeCount {
- public static void main(String[] args) throws Exception {
- Configuration conf = new Configuration();
- Job job = new Job(conf, "time_count");
- job.setOutputKeyClass(Text.class);
- job.setOutputValueClass(Text.class);
- job.setMapperClass(Map.class);
- job.setReducerClass(Reduce.class);
- job.setInputFormatClass(TextInputFormat.class);
- job.setOutputFormatClass(TextOutputFormat.class);
- FileInputFormat.addInputPath(job, new Path(args[0]));
- FileOutputFormat.setOutputPath(job, new Path(args[1]));
- job.waitForCompletion(true);
- }
- public static class Map extends Mapper<LongWritable, Text, Text, Text> {
- private Text id = new Text();
- private Text row = new Text();
- public void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {
- String line = value.toString();
- String[] items = line.split("\t");
- if (items.length == 8) {
- if (StringUtils.isNumeric(items[6])) {
- id.set(items[0] + "\t" + items[1]);
- row.set(line);
- context.write(id, row);
- }
- } else {
- System.out.println("Wrong length: " + items.length);
- }
- }
- }
- public static class Reduce extends Reducer<Text, Text, Text, Text> {
- private static final SimpleDateFormat format = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
- static {
- format.setLenient(false);
- }
- private Text rest = new Text();
- public void reduce(Text key, Iterable<Text> values, Context context)
- throws IOException, InterruptedException {
- // Parse row to Record
- ArrayList<Record> list = new ArrayList<Record>();
- for (Text row : values) {
- String[] items = row.toString().split("\t");
- try {
- Record record = new Record();
- record.items = items;
- record.start_time = format.parse(items[3]).getTime();
- record.stay_time = Long.parseLong(items[6]) * 1000;
- list.add(record);
- } catch (ParseException e) {
- e.printStackTrace();
- }
- }
- // Sort
- Collections.sort(list, new Comparator<Record>() {
- @Override
- public int compare(Record r1, Record r2) {
- return (int) (r1.start_time - r2.start_time);
- }
- });
- // Find and merge slice
- ArrayList<Record> result = new ArrayList<Record>();
- for (Record r1 : list) {
- boolean found = false;
- long r1_stop_time = r1.start_time + r1.stay_time;
- for (Record r2 : result) {
- long r2_stop_time = r2.start_time + r2.stay_time;
- if (r1.start_time > r2.start_time && r1.start_time <= r2_stop_time && r1_stop_time > r2_stop_time) {
- // merge the new slice
- r2.stay_time = r1_stop_time - r2.start_time;
- found = true;
- }
- }
- if (!found) {
- result.add(r1);
- }
- }
- // Output
- for (Record r : result) {
- key.set(r.items[0]);
- String value = r.items[1] + "\t"
- + r.items[2] + "\t"
- + r.items[3] + "\t"
- + r.items[4] + "\t"
- + r.items[5] + "\t"
- + (r.stay_time / 1000) + "\t"
- + r.items[6] + "\t";
- rest.set(value);
- context.write(key, rest);
- }
- }
- static class Record {
- String[] items;
- long start_time;
- long stay_time;
- }
- }
- }
复制代码
2.
2.1 使用 find + sed 来实现:
find /home/ocetl/app/hadoop -exec sed -i 's/\$HADOOP_HOME\$/\/home\/ocetl\/app\/hadoop/g' {} \;
2.2 直接使用ssh的参数
- #!/bin/bash
- if [ $# -ne 1 ]
- then
- echo "Usage: `basename $0` {command}"
- exit
- fi
- for i in H1 H2 H3 H4 H5 H6 H7 H8 H9 H10
- do
- echo "$i:"
- ssh $i "$1"
- done
复制代码
3.
3.1 不了解,HDFS用了缓存
3.2 问题是当硬盘空间很大,而内存页面缓存很少的时候,DN的Block report需要很长时间生成,而此时 FSVolumeSet 锁是锁住的,因此所有读写操作都无法执行,最终导致那些操作超时。此问题是建议提供一种方法使block report不需要持有FSVolumeSet锁,从而不会导致那些任务失败。
4. 只是替换分隔符从空格到逗号,以及增加搜索关键字列表:
- import org.apache.hadoop.conf.Configuration;
- 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.input.TextInputFormat;
- import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
- import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;
- import java.io.IOException;
- import java.util.ArrayList;
- public class WordCount {
- public static class Map extends Mapper<LongWritable, Text, Text, IntWritable> {
- private final static IntWritable one = new IntWritable(1);
- private Text word = new Text();
- private final static ArrayList<String> target_words = new ArrayList<String>();
- public void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {
- String[] items = value.toString().toLowerCase().replaceAll("\\p{Punct}", "").split("\\s+");
- for (String item : items) {
- if (target_words.contains(item)) {
- word.set(item);
- context.write(word, one);
- }
- }
- }
- public static void clear() {
- target_words.clear();
- }
- public static void add(String word) {
- target_words.add(word);
- }
- }
- public static class Reduce extends Reducer<Text, IntWritable, Text, IntWritable> {
- public void reduce(Text key, Iterable<IntWritable> values, Context context)
- throws IOException, InterruptedException {
- int sum = 0;
- for (IntWritable val : values) {
- sum += val.get();
- }
- context.write(key, new IntWritable(sum));
- }
- }
- public static void main(String[] args) throws Exception {
- Configuration conf = new Configuration();
- if (args.length < 3) {
- System.out.println("Usage: wordcount <input_path> <output_path> <keyword_list>");
- return;
- }
- // Add to target
- String[] target_words = args[2].split(",");
- for (String word : target_words) {
- Map.add(word.toLowerCase());
- }
- Job job = new Job(conf, "wordcount");
- job.setOutputKeyClass(Text.class);
- job.setOutputValueClass(IntWritable.class);
- job.setMapperClass(Map.class);
- job.setReducerClass(Reduce.class);
- job.setInputFormatClass(TextInputFormat.class);
- job.setOutputFormatClass(TextOutputFormat.class);
- FileInputFormat.addInputPath(job, new Path(args[0]));
- FileOutputFormat.setOutputPath(job, new Path(args[1]));
- job.waitForCompletion(true);
- }
- }
复制代码
5. 第五题的程序是什么?
6.
hosts:增加局域网主机名和ip对应关系,省得再记住ip;
hostname:该主机名,克隆虚拟机的时候经常需要这么做;
fstab:修改挂载点,加新硬盘的时候会需要;
profile, bash.bashrc: 修改系统范围环境变量时经常用;
network/interfaces:配置静态IP时需要。
7
7.1
- package org.aboutyun;
- import java.io.BufferedReader;
- import java.io.FileNotFoundException;
- import java.io.FileReader;
- import java.io.IOException;
- public class LineCounter {
- public static void main(String[] args) {
- try {
- BufferedReader reader = new BufferedReader(new FileReader(args[0]));
- char[] buffer = new char[4096];
- int count;
- char last = 0;
- long line_count = 0;
- while((count = reader.read(buffer)) >= 0) {
- if (count > 0 && line_count == 0) {
- // has something in file, so at least 1 line.
- line_count = 1;
- }
- for (int i = 0; i < count ; ++i) {
- char c = buffer[i];
- if (c == 0x02) {
- if (i == 0 && last == 0x01) {
- // buffer split the 0x01,0x02
- ++line_count;
- } else if (buffer[i-1] == 0x01) {
- // normal one
- ++line_count;
- }
- }
- }
- // keep the last one
- last = buffer[count-1];
- }
- System.out.println(line_count);
- } catch (FileNotFoundException e) {
- e.printStackTrace();
- } catch (IOException e) {
- e.printStackTrace();
- }
- }
- }
复制代码
7.2 可以使用Profiler来对性能进行评估分析,比如Eclipse的TPTP,或者JProfiler。可以观察不同函数调用次数和以及占用时间,从而减少调用次数,以及优化函数内部。
原文地址: http://www.aboutyun.com/thread-5241-1-1.html