- Hadoop Beginner's Guide
- Table of Contents
- Hadoop Beginner's Guide
- Credits
- About the Author
- About the Reviewers
- www.PacktPub.com
- Preface
- 1. What It's All About
- 2. Getting Hadoop Up and Running
- Hadoop on a local Ubuntu host
- Time for action – checking the prerequisites
- Time for action – downloading Hadoop
- Time for action – setting up SSH
- Time for action – using Hadoop to calculate Pi
- Time for action – configuring the pseudo-distributed mode
- Time for action – changing the base HDFS directory
- Time for action – formatting the NameNode
- Time for action – starting Hadoop
- Time for action – using HDFS
- Time for action – WordCount, the Hello World of MapReduce
- Using Elastic MapReduce
- Time for action – WordCount on EMR using the management console
- Comparison of local versus EMR Hadoop
- Summary
- 3. Understanding MapReduce
- Key/value pairs
- The Hadoop Java API for MapReduce
- Writing MapReduce programs
- Time for action – setting up the classpath
- Time for action – implementing WordCount
- Time for action – building a JAR file
- Time for action – running WordCount on a local Hadoop cluster
- Time for action – running WordCount on EMR
- Time for action – WordCount the easy way
- Walking through a run of WordCount
- Startup
- Splitting the input
- Task assignment
- Task startup
- Ongoing JobTracker monitoring
- Mapper input
- Mapper execution
- Mapper output and reduce input
- Partitioning
- The optional partition function
- Reducer input
- Reducer execution
- Reducer output
- Shutdown
- That's all there is to it!
- Apart from the combiner…maybe
- Time for action – WordCount with a combiner
- Time for action – fixing WordCount to work with a combiner
- Hadoop-specific data types
- Time for action – using the Writable wrapper classes
- Input/output
- Summary
- 4. Developing MapReduce Programs
- Using languages other than Java with Hadoop
- Time for action – implementing WordCount using Streaming
- Analyzing a large dataset
- Time for action – summarizing the UFO data
- Time for action – summarizing the shape data
- Time for action – correlating of sighting duration to UFO shape
- Time for action – performing the shape/time analysis from the command line
- Time for action – using ChainMapper for field validation/analysis
- Time for action – using the Distributed Cache to improve location output
- Counters, status, and other output
- Time for action – creating counters, task states, and writing log output
- Summary
- 5. Advanced MapReduce Techniques
- Simple, advanced, and in-between
- Joins
- Time for action – reduce-side join using MultipleInputs
- Graph algorithms
- Time for action – representing the graph
- Time for action – creating the source code
- Time for action – the first run
- Time for action – the second run
- Time for action – the third run
- Time for action – the fourth and last run
- Using language-independent data structures
- Time for action – getting and installing Avro
- Time for action – defining the schema
- Time for action – creating the source Avro data with Ruby
- Time for action – consuming the Avro data with Java
- Time for action – generating shape summaries in MapReduce
- Time for action – examining the output data with Ruby
- Time for action – examining the output data with Java
- Summary
- 6. When Things Break
- Failure
- Time for action – killing a DataNode process
- Time for action – the replication factor in action
- Time for action – intentionally causing missing blocks
- Time for action – killing a TaskTracker process
- Time for action – killing the JobTracker
- Time for action – killing the NameNode process
- What just happened?
- Starting a replacement NameNode
- The role of the NameNode in more detail
- File systems, files, blocks, and nodes
- The single most important piece of data in the cluster – fsimage
- DataNode startup
- Safe mode
- SecondaryNameNode
- So what to do when the NameNode process has a critical failure?
- BackupNode/CheckpointNode and NameNode HA
- Hardware failure
- Host failure
- Host corruption
- The risk of correlated failures
- Task failure due to software
- What just happened?
- Time for action – causing task failure
- Time for action – handling dirty data by using skip mode
- Summary
- 7. Keeping Things Running
- A note on EMR
- Hadoop configuration properties
- Time for action – browsing default properties
- Setting up a cluster
- Time for action – examining the default rack configuration
- Time for action – adding a rack awareness script
- Cluster access control
- Time for action – demonstrating the default security
- Managing the NameNode
- Time for action – adding an additional fsimage location
- Time for action – swapping to a new NameNode host
- Managing HDFS
- MapReduce management
- Time for action – changing job priorities and killing a job
- Scaling
- Summary
- 8. A Relational View on Data with Hive
- Overview of Hive
- Setting up Hive
- Time for action – installing Hive
- Using Hive
- Time for action – creating a table for the UFO data
- Time for action – inserting the UFO data
- Time for action – validating the table
- Time for action – redefining the table with the correct column separator
- Time for action – creating a table from an existing file
- Time for action – performing a join
- Time for action – using views
- Time for action – exporting query output
- Time for action – making a partitioned UFO sighting table
- Time for action – adding a new User Defined Function (UDF)
- Hive on Amazon Web Services
- Time for action – running UFO analysis on EMR
- Summary
- 9. Working with Relational Databases
- Common data paths
- Setting up MySQL
- Time for action – installing and setting up MySQL
- Time for action – configuring MySQL to allow remote connections
- Time for action – setting up the employee database
- Getting data into Hadoop
- Time for action – downloading and configuring Sqoop
- Time for action – exporting data from MySQL to HDFS
- Time for action – exporting data from MySQL into Hive
- Time for action – a more selective import
- Time for action – using a type mapping
- Time for action – importing data from a raw query
- Getting data out of Hadoop
- Time for action – importing data from Hadoop into MySQL
- Time for action – importing Hive data into MySQL
- Time for action – fixing the mapping and re-running the export
- AWS considerations
- Summary
- 10. Data Collection with Flume
- A note about AWS
- Data data everywhere...
- Time for action – getting web server data into Hadoop
- Introducing Apache Flume
- Time for action – installing and configuring Flume
- Time for action – capturing network traffic in a log file
- Time for action – logging to the console
- Time for action – capturing the output of a command to a flat file
- Time for action – capturing a remote file in a local flat file
- Time for action – writing network traffic onto HDFS
- Time for action – adding timestamps
- Time for action – multi level Flume networks
- Time for action – writing to multiple sinks
- The bigger picture
- Summary
- 11. Where to Go Next
- A. Pop Quiz Answers
- Index
Having a Hadoop cluster is not all about writing interesting programs to do clever data analysis. You also need to maintain the cluster, and keep it tuned and ready to do the data crunching you want.
In this chapter we will cover:
- More about Hadoop configuration properties
- How to select hardware for your cluster
- How Hadoop security works
- Managing the NameNode
- Managing HDFS
- Managing MapReduce
- Scaling the cluster
Although these topics are operationally focused, they do give us an opportunity to explore some aspects of Hadoop we have not looked at before. Therefore, even if you won't be personally managing the cluster, there should be useful information here for you too.
One of the main benefits of using cloud services such as those offered by Amazon Web Services is that much of the maintenance overhead is borne by the service provider. Elastic MapReduce can create Hadoop clusters tied to the execution of a single task (non-persistent job flows) or allow long-running clusters that can be used for multiple jobs (persistent job flows). When non-persistent job flows are used, the actual mechanics of how the underlying Hadoop cluster is configured and run are largely invisible to the user. Consequently, users employing non-persistent job flows will not need to consider many of the topics in this chapter. If you are using EMR with persistent job flows, many topics (but not all) do become relevant.
We will generally talk about local Hadoop clusters in this chapter. If you need to reconfigure a persistent job flow, use the same Hadoop properties but set them as described in Chapter 3, Writing MapReduce Jobs.
-
No Comment
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.