New Ease-of-Management Tools for HACMP
Last month's article, "Ease-of-Management Tools for Virtualization," examined maintenance issues while server consolidation and increasing workloads drive a need to maintain high availability (HA). In a perfect world, managing an HA cluster would be no harder than managing a single system with the same number of users and devices. While that happy state has not yet been achieved, the most recent releases of the IBM* High Availability Cluster Multiprocessing product (HACMP*) have provided tools and facilities that can greatly ease the HACMP cluster's administrative burden. The newer releases of HACMP, V5.1 and V5.2, are specifically designed with ease-of-use in mind.
This article is intended for those who have dealt with HACMP clusters in the past but haven't become familiar with the more recent releases. It also attempts to address the concern that HACMP clusters are always hard to set up and manage.
We'll begin our look at HACMP using the System Management Interface (SMIT) rather than command-line, although HACMP commands can be entered at command-line or, after minor configuration, an HACMP interface called WEBSMIT can be utilized. Keep in mind that although HACMP is designed to fail over systems in the event of disruptive system events, it's also a key part of maintaining availability by incorporating failover to allow for routine system maintenance such as firmware updates.
The differences in configuring a cluster with HACMP V5.1 are apparent as soon as SMIT is started: the configuration panels have been divided into Standard and Extended paths. The Standard path contains those items that will be needed in nearly every cluster. The Extended path contains items that are less commonly used or for which greater expertise is required. With this separation, the SMIT panels in the Standard path have a cleaner, simpler look. Fewer questions need be answered, and answers entered in one panel are remembered and carried forward to others.
Much of the improved ease of cluster setup comes from configuration discovery. In past releases, it was necessary to inform HACMP of the cluster's actual physical layout (e.g., what networks connected the nodes, what volume groups were shared, what interfaces are on a network). As of V5.1, HACMP discovers this physical layout and answers all of the "what connects to what" questions on its own. While this doesn't allow the cluster to be automatically configured, it does mean that HACMP can provide pick lists for many of the choices on the SMIT panels. Picking a value from a list is far easier and less error prone than manually re-entering it.
There's one case where configuration discovery allows the process of cluster setup to be even more radically simplified. If the user is building a two-node cluster and the connecting networks and shared- volume groups are already in place, then the HACMP 5.2 Two-Node Configuration wizard can be used. In this case, the user only needs to answer five questions to configure the cluster:
1. What's the name of the backup cluster node?
2. What's the name of the application to be made highly available?
3. What's the path name to start that application?
4. What's the path name to stop that application?
5. What's the IP address that the application uses?
HACMP has always provided facilities to keep its own configuration information in sync across the cluster. However, it's often also vitally important to keep application-configuration information in sync, too