Hourly web log analysis through Hadoop

Many a time one may want to parse the weblogs for doing some quick analysis on AB tests or for security/fraud alerts or recent advertisement or campaign analysis. There are many applications or utilities that perform web log analysis but more often than not regular expressions provide a powerful and elegant ways to analyse these logs and especially handy when one is dealing with massive and quickly rotating web logs.   Check out this wiki for more general info on web analytics.

When each weblog is multi-gigabytes and moved to archive every couple of hours and there is farm of web servers (hundreds or thousands of web servers), many of the vendor or 3rd party applications don't scale up either. Hadoop streaming with simple utilities can provide insights into what otherwise would have been a costly experiment.

Here is an regular expession used to extract HOUR and the string that is of interest from the Apache web logs. Each entry in the log has the format similar to the ones below.

---

01/Jun/2010:07:09:26 -0500] - 127.0.0.1 "GET /apache_pb.gif HTTP/1.1" status=200 size=2326 ab=test1AB ....

01/Jun/2010:07:09:26 -0500] - 127.0.0.1 "GET /apache_pb.gif HTTP/1.1" status=200 size=2326 ab=test-2CD ....

---

For more details on Apache log configuration, take a look at its specs and custom logs.  Along with time of the day of the site hit, user IP address, http request (GET or POST), page requested, protocol used, one can configure the web server to log many more details including referer, user agent (browser), environment variables, etc.

Adding disk space to Ubuntu (guest OS)

Recently, in one of my VMWare system (with Windows XP host OS and Ubuntu guest OS), I ran into disk space shortage while installing a package (Thrift - A cross language service).  Though there were many articles on the Internet in helping extending the guest OS partitions, they were for older versions or not applicable any more or had many complex steps.  On spending some time and digging more on the issue, I found pretty straight forward steps in achieving what I wanted to do - doubling the disk space for Ubuntu.

Initial setup:
  Host OS: Windows OS
  Guest OS: Ubuntu Karmic  (had 8G pre-allocated)
  Disk: 148G Total and 60G Free out of which 8G to be allocated to Ubuntu

Needed:
   Double Ubuntu disk space to 16G by adding a new partition

---

Step 1:

  Run "disk analyzer" and get current disk space usage data that provides total, used, available and usred.  See below.  The system has less than 10% free disk space and needs more.

---

Log Parsing through Hadoop, Hive & Python

One of the primary analysis done on web access logs is some cohort analysis where one need to pull user access date time and along with other dimensions like user, ip, geo data, etc. Here I will be using Hadoop/ Hive/ Python to pull date, ip data from access log into Hadoop and run some queries. The example illustrates using Hadoop (version 0.20.1) streaming, SERDE, Hive’s (version 0.40) plugin customer mapper (get_access_log_ip).

The steps below load few thousand rows into a target table (dw_log_ip_test – data warehouse access log) “access_log_2010_01_25” then extract date from format like DD/Mon/YYYY:HH:MM:SS -800 to ‘DD/Mon/YYYY’ along with remote ip address through a Python streaming script.

Step 1: First create a table to access log (access_log_2010_01_25) and then load data into it.

---

hive> 
CREATE TABLE access_log_2010_01_25 (
  request_date STRING,
  remote_ip STRING,
  method STRING,
  request STRING,
  protocol STRING,
  user STRING,
  status STRING,
  size STRING,
  time STRING,
  remote_host STRING,
  ts STRING,
  perf STRING
)
ROW FORMAT SERDE 'org.apache.hadoop.hive.contrib.serde2.RegexSerDe'
WITH SERDEPROPERTIES  (
"input.regex" = "\\[([^]]+)\\] ([^ ]*) \"([^ ]*) ([^ ]*) ([^ \"]*)\" user=([^ ]*) status=([^ ]*) size=([^ ]*)  time=([^ ]*) host=([^ ]*) timestamp=([^ ]*) perf=([^ ]*)",
"output.format.string" = "%1$s %2$s \"%3$s %4$s %5$s\" user=%6$s status=%7$s size=%8$s  time=%9$s  host=%10$s timestamp=%11$s  perf=%12$s"
)
STORED AS TEXTFILE;


hive> LOAD DATA LOCAL INPATH '/mnt/web_ser101/weblog_server101_20100125_1'   
    >   OVERWRITE INTO TABLE access_log_2010_01_25;
#- After load the data in one of the record would look like:
#- 25/Jan/2010:13:14:05 -0800      123.123.123.123   GET     /xmls/public/thumbnail.xml   HTTP/1.1        -       302     250     0  abcd.com   1264454045    -

Hive Metastore Derby DB

If you are using Hive in default mode, you may see the following behavior - you get to hive client from different directories and see different results when you run a query like "show tables".  For example, you have hive installed in /usr/local/hive and your are currently in your home directory and run

~> /usr/local/hive/bin/hive    #-- get to hive
hive> create table new_table (c1 string);
hive> show tables;

Now you will see "new_table" in the list.

~> cd /tmp
/tmp> /usr/local/hive/bin/hive   #-- get to hive
hive> show tables;

Now you don't see "new_table" in your list of tables.  Those who come from typical SQL background may find it little weird in the beginning due to fact that results seem different depending on from where you started the hive client.  The reason  is because hive uses "embedded Derby" database to store meta data and one of the default configuration property is to use the current directory to store metastore_db.

On starting the hive from two different directories like above, one would see two "metastore_db" directories  created in home (~) and /tmp directories.  You can change this and use one metastore_db by updating "/usr/local/hive/conf/hive-default.xml" file's "javax.jdo.option.ConnectionURL" as below.

Default setting:
    <property>
      <name>javax.jdo.option.ConnectionURL</name>
      <value>jdbc:derby:;databaseName=metastore_db;create=true</value>
      <description>JDBC connect string for a JDBC metastore</description>
    </property>

Update it to:
    <property>
      <name>javax.jdo.option.ConnectionURL</name>
      <value>jdbc:derby:;databaseName=/home/hadoop/metastore_db;create=true</value>
      <description>JDBC connect string for a JDBC metastore</description>
    </property>

"/home/hadoop" is an example and one can appropriately change it to suitable host and directory.  Say,
<value>jdbc:derby:;databaseName=//localhost:1234/metastore_db;create=true</value>

Cheers,
Shiva

Hadoop Cheat Sheet

Cheat sheet for Hadoop version 0.20.1 (Jan.2010).