用Hive分析nginx日志

这里用到的nginx日志是网站的访问日志，比如：

180.173.250.74 - - [08/Jan/2015:12:38:08 +0800] "GET /avatar/xxx.png HTTP/1.1" 200 968 
"http://www.iteblog.com/archives/994" 
"Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) 
Chrome/34.0.1847.131 Safari/537.36"

这条日志里面含有9列(为了展示的美观，我在这里面加了换行符)，每列之间是用空格分割的，每列的含义分别是客户端访问IP、用户标示、用户、访问时间、请求页面、请求状态、返回文件的大小、跳转来源、浏览器UA。如果想用一般的方法解析这条日志的话有点困难。但是如果我们会正则表达式的话，去匹配这九列数据还是很简单的：

([^ ]*) ([^ ]*) ([^ ]*) (\[.*\]) (\".*?\") (-|[0-9]*) (-|[0-9]*) (\".*?\") (\".*?\")

这样一来就可以匹配出每一列的值。而在Hive中我们是可以指定输入文件解析器(SerDe)的，并且在Hive中内置了一个org.apache.hadoop.hive.contrib.serde2.RegexSerDe正则解析器，我们可以直接使用它。所以整个建表语句可以这么写：

CREATE TABLE logs(
  host STRING,
  identity STRING,
  user STRING,
  time STRING,
  request STRING,
  status STRING,
  size STRING,
  referer STRING,
  agent STRING)
ROW FORMAT SERDE 'org.apache.hadoop.hive.contrib.serde2.RegexSerDe'
WITH SERDEPROPERTIES (
  "input.regex" = "([^ ]*) ([^ ]*) ([^ ]*) (\\[.*\\]) (\".*?\") (-|[0-9]*) 
                (-|[0-9]*) (\".*?\") (\".*?\")",
  "output.format.string" = "%1$s %2$s %3$s %4$s %5$s %6$s %7$s %8$s %9$s"
)
STORED AS TEXTFILE;

将日志放置到这个表的目录下，gz格式和未知压缩格式都可以直接被Hive解析。用下面语句可以查询出每小时访问量超过20的IP：

hive> select substring(time, 2, 14) date ,host, count(*) as count 
from logs 
group by substring(time, 2, 14), host 
having count > 20 
sort by date, count;

29/Dec/2014:00	47.18.236.106	24
29/Dec/2014:02	81.215.34.45	70
29/Dec/2014:04	66.249.64.18	23
29/Dec/2014:04	66.249.64.22	24
29/Dec/2014:09	119.145.14.213	44
29/Dec/2014:09	113.90.78.63	52
29/Dec/2014:10	106.39.255.133	26
29/Dec/2014:10	211.99.9.68	30
29/Dec/2014:10	60.10.71.97	33
29/Dec/2014:10	222.128.29.21	76
29/Dec/2014:11	91.237.69.17	56
29/Dec/2014:11	211.151.238.52	144
29/Dec/2014:12	222.92.189.35	26
29/Dec/2014:12	218.85.130.110	31
29/Dec/2014:12	218.4.189.13	77
29/Dec/2014:13	61.57.231.254	30
29/Dec/2014:13	124.207.11.123	33
29/Dec/2014:14	134.134.139.76	22
29/Dec/2014:14	218.15.33.28	27
29/Dec/2014:14	218.247.17.100	67
29/Dec/2014:15	116.235.244.139	31
29/Dec/2014:15	101.231.119.202	52
29/Dec/2014:15	183.11.249.158	64
29/Dec/2014:16	116.235.244.139	22
29/Dec/2014:16	211.151.238.52	30
29/Dec/2014:16	123.138.184.84	53
29/Dec/2014:17	219.159.77.110	55
29/Dec/2014:17	87.204.102.195	57
29/Dec/2014:17	111.203.3.1	77
29/Dec/2014:18	125.41.147.243	21
29/Dec/2014:18	66.249.64.18	23
29/Dec/2014:18	101.251.230.3	39
29/Dec/2014:18	110.249.70.182	40
29/Dec/2014:18	91.200.12.26	44
29/Dec/2014:18	218.64.17.230	93
29/Dec/2014:19	66.249.64.22	27
29/Dec/2014:21	222.129.35.102	25

或者是一些其他的操作。

如果你对Bash比较熟悉的话，那你完全可以不用到Hive，直接用awk、sort等函数来实现，比如我想知道今天访问量比较多的IP并对他们进行排序，取前10条的语句可以这么写：

[root@iteblog ]# awk '{print $1}' www.iteblog.com.access.log | sort | uniq -c |
> sort -nr | head -n 10
    241 46.119.121.149
    224 66.249.65.51
    220 66.249.65.49
    219 66.249.65.47
    210 211.151.238.52
    184 207.46.13.96
    183 157.55.39.44
    182 112.247.104.147
    173 157.55.39.239
    169 157.55.39.106

文章来自 过往记忆： http://www.iteblog.com/archives/1250