Handling of resource groups configured with IPAT via IP aliases

IP aliasing is a function of the TCP/IP stack where multiple IP addresses can be added to the same physical interface. PowerHA SystemMirror uses IP aliases for recovery, so the base address for the interface does not change. PowerHA SystemMirror recovers the service address by adding it as a second, or alias address, on the same interface. A single physical interface can host or back up multiple service addresses. This greatly improves the configuration flexibility and fallover options by requiring fewer physical resources to serve as backup. IPAT via IP Aliases is also faster as there are fewer commands required when moving addresses.

To control the placement of the service IP label aliases on the cluster node physical network interface cards, you can configure a distribution preference for the aliases of the service IP labels that are placed under PowerHA SystemMirror control.

Resource group behavior when using IPAT via IP aliases

With IPAT via IP Aliases, the service address is added as an alias address on an available boot interface. This applies to the node where the resource group is first acquired, as well as to the node(s) that might acquire it later. When the resource group is released, the service address is removed from the interface, but this does not alter the base or boot address on the interface.

The mechanics of IP address takeover on a network using aliases works the same way for all non-concurrent resource groups. While the mechanics are identical, IPAT via IP aliases does affect the initial startup and fallover placement of the resource group.

The aliased service IP labels are distributed across all available boot interfaces. To facilitate even distribution of labels across all available IP interface cards, PowerHA SystemMirror sorts all available interfaces by state and then by the number of aliased addresses already placed on the interface, and places the aliased labels accordingly. Note that this distribution is only done at fallover time, PowerHA SystemMirror makes no attempt to redistribute the labels later, if another interface becomes active.

Resource group placement on cluster startup

The presence of a service IP label in the resource group does not change the placement policy for the resource group on initial cluster startup. Therefore, on initial cluster startup, a non-concurrent resource group is placed according to the defined startup policy.

On the subsequent cluster startup, PowerHA SystemMirror moves the resource group containing the service IP label onto a node with a boot interface that:

• Is up
• Has a different subnet than the IP label that is being moved.

In addition, PowerHA SystemMirror follows these rules:

• If multiple boot interfaces are found that are up and have different subnets, then PowerHA SystemMirror moves the resource group onto the one that comes first in the alphabetically-sorted list of network interfaces configured on the node.
• If the resource group uses Online Using Node Distribution Policy startup, the resource group is placed on a node that does not host another resource group.

Resource group placement on fallover

On fallover, if you have configured resource groups that contain aliased service IP labels, this allows having more than one non-concurrent resource group on the same node. Therefore, more than one resource group can be serviced by a node with a single physical interface.

On fallover, PowerHA SystemMirror moves the resource group containing the service IP label onto a node with a boot interface which:

• Is up
• Has a different subnet
  • Is preferably not hosting another service label (if available).
  • Comes first in the alphabetically-sorted list of network interfaces in the network configuration.

The key advantage of IPAT via IP Aliases is that, on fallover, more than one resource group can be serviced by a node with a single physical interface.