A partition is a group of managed devices formed by IP Availability Manager to help in root-cause analysis. IP Availability Manager builds a topology that describes managed devices in your network and their interconnections. If a device is modeled in the topology, there must be a path in the actual network that connects the device to the management station. Ideally, IP Availability Manager can also have representations of all the devices on that path.
However, some devices along the path might not actually be represented in your network topology. For example, those devices might not have SNMPAgents, or IP Availability Manager might not have access to their SNMPAgents, or access to the agents might be blocked by a firewall or some other administrative mechanism. Also, if you do not use autodiscovery, your seed file or information obtained from another Network Management System (NMS) might be incomplete. Or, you might have explicitly unmanaged or deleted some of the devices which are on that path.
IP Availability Manager's root-cause analysis algorithms need to determine the connectivity among managed devices. When the topology is incomplete, IP Availability Manager creates partitions to group all the related managed devices, which helps analysis to proceed.
Within a single partition, any two managed devices are indeed connected by a path. However, objects within distinct partitions have no path between them in the modeled topology, even though they are connected in the actual network.
Because many configurations produce large numbers of apparently isolated objects, usually hosts, IP Availability Manager only creates partitions with two or more devices.
You can assign names to the partitions in your network using the partition.conf file.