Book Image

Hadoop 2.x Administration Cookbook

By : Aman Singh
Book Image

Hadoop 2.x Administration Cookbook

By: Aman Singh

Overview of this book

Hadoop enables the distributed storage and processing of large datasets across clusters of computers. Learning how to administer Hadoop is crucial to exploit its unique features. With this book, you will be able to overcome common problems encountered in Hadoop administration. The book begins with laying the foundation by showing you the steps needed to set up a Hadoop cluster and its various nodes. You will get a better understanding of how to maintain Hadoop cluster, especially on the HDFS layer and using YARN and MapReduce. Further on, you will explore durability and high availability of a Hadoop cluster. You’ll get a better understanding of the schedulers in Hadoop and how to configure and use them for your tasks. You will also get hands-on experience with the backup and recovery options and the performance tuning aspects of Hadoop. Finally, you will get a better understanding of troubleshooting, diagnostics, and best practices in Hadoop administration. By the end of this book, you will have a proper understanding of working with Hadoop clusters and will also be able to secure, encrypt it, and configure auditing for your Hadoop clusters.
Table of Contents (20 chapters)
Hadoop 2.x Administration Cookbook
Credits
About the Author
About the Reviewers
www.PacktPub.com
Customer Feedback
Preface
Index

Configuring storage based policies


Over a period of time, the Hadoop clusters grow in size and the hardware will change. More powerful nodes, with faster CPU, disks, and RAM, will be added and there will be an uneven performance of nodes across the cluster.

This can be controlled at the node level using YARN labels, which we will see at a later stage in the book. We can also configure storage-based policies according to the disk drives we have such as SSDs, SATA, and PATA.

Getting ready

For this recipe, you will again need a running cluster with at least HDFS configured and Datanodes in the cluster with different types of hard drives. Some nodes have slow disk drives and others have fast disk drives.

The segregation can also be based on the cost factor, as SSDs are expensive and must be on nodes, which are SLA stringent. The classification can be done into cold storage for the archiving of data and the hot storage of live data in the cluster. Users can store performance-critical data in SSD...