Administration Guide

Cluster Components

Each cluster consists of the following components and resources:

Physical machine

Each physical machine has one public network interface, one or more private network interfaces on the public network, a set of shared disks, and a disk for the operating system. Each cluster can contain up to four physical machines.

Logical host

The logical host essentially borrows the CPU or (CPUs) and memory from the physical machine, and migrates from machine to machine during a failover situation. Each logical host consists of the following resources:

• One logical interface on the public network with its own IP address and hostname

  Remote clients should always use this IP address when connecting to failover services, because the address is moved from one machine to another machine during a failover.

• One or more disk groups

  The disk group (or disk set) is a collection of physical disks that are associated with the logical host. The disk group must be on disks that are physically shared between the two machines in a cluster.

• One high availability service (that is, DB2)

  The high availability service provides a set of scripts that Sun Cluster 2.1 can use to start, stop, and abort the service.

  When a failover occurs, the high availability service from one machine fails over to another machine. You must ensure that the physical machine has enough CPU and memory resource to properly run the system after the failover; otherwise, the services may fail.

You can have as many logical hosts as you want on a machine, but for administrative reasons, it is recommended that you assign no more than one to a machine. The following is an example of the layout for a logical host filesystem for Sun Cluster 2.1 with DB2 (See the Sun Cluster 2.1 documentation for instructions on how to add logical host filesystems). The name of the logical host in this example is snap:

/snap/

The logical host filesystem (needed for Sun Cluster 2.1).

/snap/home/

The place to put the high availability instance home directory.

/snap/disks/

The place where SMS filesystems should be placed to make sure that are available after a failover.

Private network

Private networks are used for communicating between two nodes. Heartbeat messages and Remote Procedure Calls (RPCs) travel over these networks to keep the two nodes in synchronous operation, so that they can back up each other in the event of a failover.

Public networks

The public network includes all the primary and logical network interfaces and IP addresses. The logical network interfaces or logical hosts should be referred to when communicating with DB2 on the cluster because these are failed over to the remaining machine at the time of the failover.

Disk group

Disk groups contain one or more shared disks and a list of hosts that can access these disks. Only one host can own the disk groups for exclusive use at a time.

Disk mirroring

It is highly recommended that you mirror disks to increase disk availability. Without mirroring, there is a single point of failure for each disk in the system.

Figure 85 shows an example of the components in a cluster.

Figure 85. Components in a Cluster

The following sections describe the different types of failover support, and how to implement them.

[ DB2 List of Books | Search the DB2 Books ]