Cluster Terminology

The list below includes basic terms used in the HACMP environment.
Note:In general, terminology for HACMP is based on industry conventions for high availability. However, the meaning of some of the terms in HACMP may differ from the generic terms.
An application is a service, such as a database, or a collection of system services and their dependent resources, such as a service IP label and application’s start and stop scripts, that you want to keep highly available with the use of HACMP.
An application server is a collection of application start and stop scripts that you provide to HACMP by entering the pathnames for the scripts in the WebSMIT user interface. An application server becomes a resource associated with an application, you include it in a resource group for HACMP to keep it highly available. HACMP ensures that the application can start and stop successfully no matter on which cluster node it is being started.
A cluster node is a physical machine, typically an AIX or a Linux server on which you install HACMP. A cluster node also hosts an application. A cluster node serves as a server for application’s clients. HACMP’s role is to ensure continuous access to the application, no matter on which node in the cluster the application is currently active.
A home node is a node on which the application is hosted, based on your default configuration for the application’s resource group, and under normal conditions.
A takeover node is a backup cluster node to which HACMP may move the application. You can move the application to this node manually, for instance, to free the home node for planned maintenance. Or, HACMP moves the application automatically, due to a cluster component failure.
In HACMP for Linux v.5.4.1, a cluster configuration includes up to eight nodes. Therefore, you can have more than one potential takeover nodes for a particular application. You define the list of nodes on which you want HACMP to host your application using the WebSMIT interface. This list is called a resource group’s nodelist.
A cluster IP network is used for cluster communications between the nodes and for sending heartbeating information. All IP labels configured on the same HACMP network share the netmask, but may be required to have different subnets.
An IP label is a name of a network interface card (NIC) that you provide to HACMP. Network configuration for HACMP requires planning for several types of IP labels:
•Base (or boot) IP labels on each node—the ones through which an initial cluster connectivity is established.
•Service IP labels for each application—the ones through which a connection for a highly available application is established.
•Backup IP labels (optional).
•Persistent IP labels on each node. These are node-bound IP labels that are useful to have in the cluster for administrative purposes.

Note that to ensure high availability and access to the application, HACMP “recovers” the service IP address associated with the application on another node in the cluster in cases of network interface failures. HACMP uses IP aliases for HACMP networks. For information, see Planning IP Networks and Network Interfaces.
An IP alias is an alias placed on an IP label. It coexists on an interface along with the IP label. Networks that support Gratuitous ARP cache updates enable configuration of IP aliases.
IP Address Takeover (IPAT) is a process whereby a service IP label on one node is taken over by a backup node in the cluster. HACMP uses IPAT to provide high availability of IP service labels that belong to resource groups. These labels provide access to applications. HACMP uses IPAT to recover the IP label on the same node or the backup node. HACMP for Linux by default supports the mode of IPAT known as IPAT via IP Aliasing. (The other method of IPAT—IPAT via IP Replacement is not supported).
IP Address Takeover via IP Aliasing is the default method of IPAT used in HACMP. HACMP uses IPAT via IP Aliasing in cases when it must automatically recover a service IP label on another node. To configure IPAT via IP Aliasing, you configure service IP labels and their aliases to the system. When HACMP performs IPAT during automatic cluster events, it places an IP alias recovered from the “failed” node on top of the service IP address on the takeover node. As a result, access to the application continues to be provided.
Cluster resources can include an application server and a service IP label. All or some of these resources can be associated with an application you plan to keep highly available. You include cluster resources into resource groups.
A resource group is a collection of cluster resources.
Resource group startup is an activation of a resource group and its associated resources on a specified cluster node. You choose a resource group startup policy from a predefined list in WebSMIT.
Resource group fallover is an action of a resource group, when HACMP moves it from one node to another. In other words, a resource group and its associated application fall over to another node. You choose a resource group fallover policy from a predefined list in WebSMIT.
Takeover is an automatic action during which HACMP takes over resources from one node and moves them to another node. Takeover occurs when a resource group falls over to another node. A backup node is referred to as a takeover node.
Resource group fallback is an action of a resource group, when HACMP returns it from a takeover node back to the home node. You choose a resource group fallback policy from a predefined list in WebSMIT.
Cluster Startup is the starting of HACMP cluster services on the node(s).
Cluster Shutdown is the stopping of HACMP cluster services on the node(s).
Pre- and post-events are customized scripts provided by you (or other system administrators), which you can make known to HACMP and which will be run before or after a particular cluster event. For more information on pre- and post-event scripts, see the chapter on Planning Cluster Events in the HACMP for AIX Planning Guide.