View figure.
Planning your HACWS network configuration is a complex task which requires understanding the basic HACMP concepts. These concepts are explained in the HACMP publications. This section demonstrates how to plan your HACWS network configuration through a hypothetical situation. Additional specific HACWS network requirements are also described in this section.
Assume that your system has a single control workstation named dutchess.xyz.com and it will serve as the primary control workstation after you install HACWS. The workstation you add will become the backup control workstation. The name of the backup control workstation is ulster.xyz.com.
The SP nodes get control workstation services by accessing the network interface whose name matches the host name of the primary control workstation. In this example, the SP nodes get control workstation services by accessing dutchess.xyz.com. If the primary control workstation fails and the backup control workstation takes over, the backup control workstation assumes the network identity of dutchess.xyz.com.
The dutchess.xyz.com network interface gets configured on the control workstation currently providing the control workstation services. HACMP refers to dutchess.xyz.com as a service address (or service interface). The primary control workstation must use a different network address when it reboots in order to avoid a network address conflict between the two control workstations. HACMP refers to this alternate network address as a boot address (or boot interface). In this example, the boot address of the primary control workstation is dutchess_bt.xyz.com.
In addition, HACWS requires that the backup control workstation must always be reachable via a network interface whose name matches its host name. In this example, this name is ulster.xyz.com. This network interface does not get identified to HACMP. If you have no available adapter upon which to configure the ulster.xyz.com network interface, you can use an IP address alias.
Each control workstation in this example configuration contains one Ethernet adapter, connected to the SP Ethernet network. After the two control workstations are booted and before HACMP is started, their network configuration looks like the one illustrated in Figure 20.
Figure 20. Initial control workstation network configuration
View figure.
At this point, neither machine is providing control workstation services, so the dutchess.xyz.com network interface is not available. The Ethernet adapter on the primary is configured with its boot address dutchess_bt.xyz.com and the Ethernet adapter on the backup is configured with its boot address ulster_bt.xyz.com. Since there is only one network adapter, the network interface ulster.xyz.com must be configured as an IP address alias on the backup control workstation.
When the operator starts HACMP on both control workstations, the first control workstation to start HACMP becomes the active control workstation. (The operator selects the machine to become the active control workstation by starting HACMP on it first.) If HACMP is first started on the primary control workstation and then on the backup control workstation, the network configuration looks like the one illustrated in Figure 21.
Figure 21. Starting HACMP
View figure.
The only change to the network configuration is that the boot address dutchess_bt.xyz.com on the primary control workstation has been replaced by the service address dutchess.xyz.com.
If the primary control workstation should fail and the backup control workstation take over, the network interface looks like the one illustrated in Figure 22.
Figure 22. Control workstation failover
View figure.
If the primary control workstation is still running, then its Ethernet adapter is back on its boot address dutchess_bt.xyz.com, and the boot address ulster_bt.xyz.com on the backup control workstation has been replaced by the service address dutchess.xyz.com. The SP nodes continue to get control workstation services by accessing dutchess.xyz.com.
You can identify multiple network interfaces to move back and forth between the two control workstations along with the control workstation services. Some possible reasons for doing this are:
Each of these network interfaces is effectively a service address. However, the number of service addresses identified to HACMP cannot exceed the number of network adapters. Use IP address aliases to make up the difference.
In this example, each control workstation has only one network adapter. Since dutchess.xyz.com is defined to HACMP as a service address, any additional "effective" service addresses must be configured using IP address aliases. If you added an SP system partition whose network interface name on the control workstation is columbia.xyz.com to this example configuration, it would look like Figure 23 when the backup control workstation is active.
Figure 23. Adding an SP system partition
View figure.
The HACMP service address dutchess.xyz.com is configured on adapter en0 on the backup control workstation and the network interfaces columbia.xyz.com and ulster.xyz.com are configured on adapter en0 as IP address aliases. The service address dutchess.xyz.com is identified to HACMP. For each service address that is identified to HACMP, there must be boot addresses for both control workstations. The boot address dutchess_bt.xyz.com is identified to HACMP for the primary control workstation, and the boot address ulster_bt.xyz.com is identified to HACMP for the backup control workstation
At this point, if you have not done so already, you need to do the following: