The Sun Cluster course I attended at Sun Microsystems was interesting, but it left me wondering how one would actually go about building cluster applications. This book describes how to do it.

It begins with a chapter in which the various cluster architectures (high availability, server farm, high performance, and grid) are described. The concept of application availability is discussed, and some sample scenarios are presented that show the cost considerations for various levels of availability.

In chapter 2, the particular capabilities of Sun Cluster 3 are introduced. Global devices and global disk identifiers are described, and line diagrams are used to show how global filesystems can be accessed by nodes that are cluster members. The use of failover filesystems as an alternative to global filesystems is also illustrated. The creation of cluster resources and cluster resource groups is covered with command-line examples, and the role and realization of quorum devices is discussed.

Chapter 3 takes the reader through the process of preparing a global filesystem using Solaris Volume Manager (previously known as Solstice DiskSuite) with soft partitions on a mirrored volume, and then installing the Sun Apache Web server packages and modifying its configuration for cluster use. A generic data service (GDS) agent is then installed and activated to monitor the Web server application. All necessary commands are given in full with explanations—really useful stuff!

The importance of carefully planning a cluster application is covered in chapter 4. Some key issues here are setting the bounds on recovery times, installation directories for executables (locally or on shared storage), configuration of logical interfaces, and application monitoring.

Chapter 5 is about installing and using the SunPlex Agent Builder. Both graphical user interface (GUI) and command-line procedures are covered, with GDS, Korn shell, and C language options. The end product is an agent package, and the text illustrates how this can be installed like any other Sun package using the pkgadd tool.

The remaining eight chapters are, in the words of the authors, “geared toward more experienced systems administrators or application builders.” The first of these chapters deals with the resource group manager. It includes block diagrams illustrating resource start, stop, and monitor concepts; a sample resource type registration file is listed and used in examples showing how failover and other activities can be initiated. The algorithm used for Oracle database monitoring is outlined as an example of a real-world application.

Subsequent chapters cover programming with the resource management application programming interface (API), and process management. The latter includes copious examples showing how timeouts, retry intervals, and action programs can be defined. There is also a chapter about using the data service development library (DSDL); it is quite comprehensive, but you will need a good grasp of C to understand it.

I was impressed by the extensive coverage of development for existing applications. A Korn shell resource is used for a MySQL example, and a C resource (using the DSDL) is used for a common Unix printing system (CUPS) application.

Scalable services can be created within the Sun Cluster 3 framework, and can be used to dynamically improve system response times and reduce recovery times. A chapter is included illustrating some of the applicable techniques.

The final two chapters deal with the use of remote shared memory (RSM) and the development of cluster-aware applications. The RSM examples are quite comprehensive, with code segments included to check for errors, such as failure to import requested memory segments.

Much of the information in this book can be found in other Sun publications, but you would still have to work out how to use it. So why would you bother? This book has it all, complete with realistic examples.