Highly Available WebSphere Business Integration Solutions

88 Highly Available WebSphere Business Integration Solutions

6.1 Introduction

This chapter provides samples to help the reader get started contemplating their

test plans. It is outside the scope of this redbook to provide a treatise on test

methodology and the like. We all know that proper, thorough testing of our

applications, systems, environments, etc. is critical to ensuring as painless a

transition as possible when new elements are promoted into the production

environment.

So, briefly, as in any testing start with the basics: Convince yourself that the HA

cluster is working and configured as you expect. Can you ping one machine in

the HA cluster from another machine in the cluster? What about a machine

outside the cluster? Work your way slowly through the expected functionality.

The tests should build on one another and in complexity as you go along. In

considering HA in particular, the authors found it helpful to keep the following

goals in mind:

򐂰 Most importantly, the applications (flows, collabs, etc.) can recover and

restart.

򐂰 The applications will continue to process or recover when a component fails.

򐂰 The applications will perform under heavy load.

HA is also about performance of the application when individual components

are under stress. Not just failure.

򐂰 The scenario works when there are multiple Message Brokers and

InterChange Servers; for example, a flow fails on one node and is restarted

on another.

6.2 Verify basic machine failover

Both nodes are up.

1. Pull the plug

a. Pull plug on node 1.

b. Does HACMP failover occur?

c. Repeat for node 2.

2. Ping from outside the HA cluster:

a. Ping node 1 from an AIX box that is not part of the HA cluster.

Note: Throughout this chapter, the term kill refers to gracefully stopping

the component using available facilities, or stopping the component

ungracefully by terminating its processes. Either is equally acceptable.