Logs come in all shapes, but as applications and infrastructures grow, the result is a massive amount of distributed data that's useful to mine. From web and mail servers to kernel and boot logs, modern servers hold a rich set of information. Massive amounts of distributed data are a perfect application for Apache Hadoop, as are log files—time-ordered structured textual data.
You can use log processing to extract a variety of information. One of its most common uses is to extract errors or count the occurrence of some event within a system (such as login failures). You can also extract some types of performance data, such as connections or transactions per second. Other useful information includes the extraction (map) and construction of site visits (reduce) from a web log. This analysis can also support detection of unique user visits in addition to file access statistics.
These exercises give you practice in:
- Getting a simple Hadoop environment up and running
- Interacting with the Hadoop file system (HDFS)
- Writing a simple MapReduce application
- Writing a filtering Apache Pig query
- Writing an accumulating Pig query
To get the most from these exercises, you should have a basic working knowledge of Linux®. Some knowledge of virtual appliances is also useful for bringing a simple environment up.
There are two ways to get Hadoop up and running. The first is to install the Hadoop software, and then configure it for your environment (the simplest case is a single-node instance, in which all daemons run in a single node). See Distributed data processing with Hadoop, Part 1: Getting started for details.
The second and simpler way is through the use of the Cloudera's Hadoop Demo VM (which contains a Linux image plus a preconfigured Hadoop instance). The Cloudera virtual machine (VM) runs on VMware, Kernel-based Virtual Machine (KVM), or Virtualbox.
Choose a method, and complete the installation. Then, complete the following task:
- Verify that Hadoop is running by issuing an HDFS
The HDFS is a special-purpose file system that manages data and replicas within a
Hadoop cluster, distributing them to compute nodes for efficient processing.
Even though HDFS is a special-purpose file system, it implements many of the
typical file system commands. To retrieve help information for
Hadoop, issue the command
hadoop dfs. Perform
the following tasks:
- Create a test subdirectory within the HDFS.
- Move a file from the local file system into the HDFS subdirectory using
- For extra credit, view the file within HDFS using a
As demonstrated in Distributed data processing with Hadoop, Part 3: Application development, writing a word count map and reduce application is simple. Using the Ruby example demonstrated in this article, develop a Python map and reduce application, and run them on a sample set of data. Recall that Hadoop sorts the output of map so that like words are contiguous, which provides a useful optimization for the reducer.
As you saw in Data processing with Apache Pig, Pig allows you to build simple scripts that are translated into MapReduce applications. In this exercise, you extract all log entries (from /var/log/messages) that contain both the word kernel: and the word terminating.
- Create a script that extracts all log lines with the predefined criteria.
Log messages are generated by a variety of sources within the Linux kernel (such as
dhclient). In this
example, you want to discover the various sources that generate log messages and
the number of log messages per source.
- Create a script that counts the number of log messages for each log source.
The specific output depends on your particular Hadoop installation and configuration.
Listing 1. Performing an ls operation on the HDFS
$ hadoop dfs -ls / drwxrwxrwx - hue supergroup 0 2011-12-10 06:56 /tmp drwxr-xr-x - hue supergroup 0 2011-12-08 05:20 /user drwxr-xr-x - mapred supergroup 0 2011-12-08 10:06 /var $
More or fewer files might be present depending on use.
In Exercise 2, you create a subdirectory within HDFS and copy a file into it. Note
that you create test data by moving the kernel message buffer into a file. For
extra credit, view the file within the HDFS using the
command (see Listing 2).
Listing 2. Manipulating the HDFS
$ dmesg > kerndata $ hadoop dfs -mkdir /test $ hadoop dfs -ls /test $ hadoop dfs -copyFromLocal kerndata /test/mydata $ hadoop dfs -cat /test/mydata Linux version 2.6.18-274-7.1.el5 (firstname.lastname@example.org)... ... e1000: eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: RX $
In Exercise 3, you create a simple word count MapReduce application in Python. Python is actually a great language in which to implement the word count example. You can find a useful writeup on Python MapReduce in Writing a Hadoop MapReduce Program in Python by Michael G. Noll.
This example assumes that you performed the steps of exercise 2 (to ingest data into the HDFS). Listing 3 provides the map application.
Listing 3. Map application in Python
#!/usr/bin/env python import sys for line in sys.stdin: line = line.strip() words = line.split() for word in words: print '%s\t1' % word
Listing 4 provides the reduce application.
Listing 4. The reduce application in Python
#!/usr/bin/env python from operator import itemgetter import sys last_word = None last_count = 0 cur_word = None for line in sys.stdin: line = line.strip() cur_word, count = line.split('\t', 1) count = int(count) if last_word == cur_word: last_count += count else: if last_word: print '%s\t%s' % (last_word, last_count) last_count = count last_word = cur_word if last_word == cur_word: print '%s\t%s' % (last_word, last_count)
Listing 5 illustrates the process of invoking the Python MapReduce example in Hadoop.
Listing 5. Testing Python MapReduce with Hadoop
$ hadoop jar /usr/lib/hadoop-0.20/contrib/streaming/hadoop-streaming-0.20.2-cdh3u2.jar \ -file pymap.py -mapper pymap.py -file pyreduce.py -reducer pyreduce.py \ -input /test/mydata -output /test/output ... $ hadoop dfs -cat /test/output/part-00000 ... write 3 write-combining 2 wrong. 1 your 2 zone: 2 zonelists. 1 $
In Exercise 4, you extract /var/log/messages log entries that contain both the word kernel: and the word terminating. In this case, you use Pig in local mode to query the local file (see Listing 6). Load the file into a Pig relation (log), filter its contents to only kernel messages, and then filter that resulting relation for terminating messages.
Listing 6. Extracting all kernel + terminating log messages
$ pig -x local grunt> log = LOAD '/var/log/messages'; grunt> logkern = FILTER log BY $0 MATCHES '.*kernel:.*'; grunt> logkernterm = FILTER logkern BY $0 MATCHES '.*terminating.*'; grunt> dump logkernterm ... (Dec 8 11:08:48 localhost kernel: Kernel log daemon terminating.) grunt>
In Exercise 5, extract the log sources and log message counts from /var/log/messages.
In this case, create a script for the query, and execute it through Pig's local mode.
In Listing 7, you load the file and parse the input using a space
as a delimiter. You then assign the delimited string fields to your named elements.
GROUP operator to group the messages by their
source, and then use the
FOREACH operator and
COUNT to aggregate your data.
Listing 7. Log sources and counts script for /var/log/messages
log = LOAD '/var/log/messages' USING PigStorage(' ') AS (month:chararray, \ day:int, time:chararray, host:chararray, source:chararray); sources = GROUP log BY source; counts = FOREACH sources GENERATE group, COUNT(log); dump counts;
The result is shown executed in Listing 8.
Listing 8. Executing your log sources script
$ pig -x local logsources.pig ... (init:,1) (gconfd,12) (kernel:,505) (syslogd,2) (dhclient:,91) (localhost,1168) (gpm:,2) [gpm:,2) (NetworkManager:,292) (avahi-daemon:,37) (avahi-daemon:,44) (nm-system-settings:,8) $
- Distributed computing with Linux and Hadoop (Ken Mann and M. Tim Jones, developerWorks, December 2008): Discover Apache's Hadoop, a Linux-based software framework that enables distributed manipulation of vast amounts of data, including parallel indexing of internet web pages.
- Distributed data processing with Hadoop, Part 1: Getting started (M. Tim Jones, developerWorks, May 2010): Explore the Hadoop framework, including its fundamental elements, such as the Hadoop file system (HDFS), common node types, and ways to monitor and manage Hadoop using its core web interfaces. Learn to install and configure a single-node Hadoop cluster, and delve into the MapReduce application.
- Distributed data processing with Hadoop, Part 2: Going further (M. Tim Jones, developerWorks, June 2010): Configure a more advanced setup with Hadoop in a multi-node cluster for parallel processing. You'll work with MapReduce functionality in a parallel environment and explore command line and web-based management aspects of Hadoop.
- Distributed data processing with Hadoop, Part 3: Application development (M. Tim Jones, developerWorks, July 2010): Explore the Hadoop APIs and data flow and learn to use them with a simple mapper and reducer application.
- Data processing with Apache Pig (M. Tim Jones, developerWorks, February 2012): Pigs are known for rooting around and digging out anything they can consume. Apache Pig does the same thing for big data. Learn more about this tool and how to put it to work in your applications.
- Writing a Hadoop MapReduce Program in Python (Michael G. Noll, updated October 2011, published September 2007): Learn to write a simple MapReduce program for Hadoop in the Python programming language in this tutorial.
- The Open Source developerWorks zone provides a wealth of information on open source tools and using open source technologies.
- developerWorks Web development specializes in articles covering various web-based solutions.
- Stay current with developerWorks technical events and webcasts focused on a variety of IBM products and IT industry topics.
- Attend a free developerWorks Live! briefing to get up-to-speed quickly on IBM products and tools, as well as IT industry trends.
- Watch developerWorks on-demand demos ranging from product installation and setup demos for beginners, to advanced functionality for experienced developers.
- Follow developerWorks on Twitter, or subscribe to a feed of Linux tweets on developerWorks.
Get products and technologies
- Cloudera's Hadoop Demo VM (May 2012): Start using with Apache Hadoop with a set of virtual machines that include a Linux image and a preconfigured Hadoop instance.
- Evaluate IBM products in the way that suits you best: Download a product trial, try a product online, use a product in a cloud environment, or spend a few hours in the SOA Sandbox learning how to implement Service Oriented Architecture efficiently.
- Check out developerWorks blogs and get involved in the developerWorks community.
- Get involved in the developerWorks community. Connect with other developerWorks users while exploring the developer-driven blogs, forums, groups, and wikis.
M. Tim Jones is an embedded firmware architect and the author of Artificial Intelligence: A Systems Approach, GNU/Linux Application Programming (now in its second edition), AI Application Programming (in its second edition), and BSD Sockets Programming from a Multilanguage Perspective. His engineering background ranges from the development of kernels for geosynchronous spacecraft to embedded systems architecture and networking protocols development. Tim is a platform architect with Intel and author in Longmont, Colo.