Network interface events
The cluster manager reacts to the failure, unavailability, or joining of network interfaces by initiating an event.
The following table shows the network interface events.
| Network interface events | Event description |
|---|---|
| swap_adapter | This event occurs when the interface that hosts a service IP
label on a node fails. The swap_adapter event moves the service
IP label onto a boot interface on the same PowerHA® SystemMirror® network and
then reconstructs the routing table. If the service IP label is an
IP alias, it is put onto the boot interface as an additional IP label.
Otherwise, the boot IP label is removed from the interface and placed
on the failed interface. If the interface now holding the service
IP label later fails, swap_adapter can switch to another boot
interface if one exists. If a persistent node IP label was assigned
to the failed interface, it moves with the service label to the boot
interface. Note: PowerHA SystemMirror removes
IP aliases from interfaces at shutdown. It creates the aliases again
when the network becomes operational. The hacmp.out file records
these changes. |
| swap_adapter_complete | This event occurs only after a swap_adapter event has been successfully completed. The swap_adapter_complete event ensures that the local address resolution protocol (ARP) cache is updated by deleting entries and pinging cluster IP addresses. |
| fail_interface | This event occurs if an interface fails and there is no boot interface available to recover the service address. Takeover service addresses are monitored. It is possible to have an interface failure and no available interface for recovery and another interface up on the same network. This event applies to all networks, including those using IP aliasing for recovery. When a boot NIC fails on a network that is configured for IPAT via IP aliases, the fail_interface event is run. If the interface that failed was a service label, an rg_move event is then triggered. |
| join_interface | This event occurs if a boot interface becomes available or recovers. This event applies to all networks, including those using IPAT via IP aliases for recovery. Networks using IP aliases by definition do not have boot interfaces defined, so the join_interface event that is run in this case simply indicates that a boot interface joins the cluster. |
Failure of a single network interface does not generate events
If only one network interface is active on a network, the cluster manager cannot generate a failure event for that network interface, because it has no peers with which to communicate for determining the health of the interface. Situations that have a single network interface follow:
- One-node clusters
- Multinode clusters with only one node active
- Multinode clusters with virtual Ethernet interfaces
- Failure of all but one interface on a network, one at a time
For example, starting a cluster with all service or boot interfaces disconnected produces the following results:
- First node active: No failure events are generated.
- Second node active: One failure event is generated.
- Third node active: One failure event is generated.