UniInt phase 2 failover: miscellaneous scenarios

The following Phase 2 failover scenarios could potentially affect whether the interface continues to collect and buffer data.

Scenario 1: Access lost to shared file and PI server

This scenario occurs for example, if the NIC goes bad, router/switch goes down, cable is unplugged, etc. This scenario has two points of failure. Uniint failover is designed to function correctly for a single point of failure in the system. If an interface loses communication with the shared file and the PI server, the interface instance will remain in its current state. The interface will not assume primary if the interface was in backup mode at the time that both connections went down, and similarly the interface will not transition out of primary if the interface was in primary mode when both connections went down.

Scenario 2: PI Server Collective architecture

This scenario occurs when primary PI server fails and shared file access fails. Each instance reads from its designated "/host" and receives outputs from its designated "/host"; therefore, the interface instance will remain in its current state.

Scenario 3: Disaster recovery

In this scenario, the shared file is on the interface node which has a system failure that requires the entire node to be rebuilt. In this situation, the interface will failover to the other node and collect/buffer data. When the failed node is rebuilt, the interface will get back in sync automatically once the file is created. The file will be created when the interface on the failed node is started, or if the sync file was backed up on the ghost image. Note that if the sync file was on a separate machine from the interface nodes and that machine crashed, the sync node must be restored from a ghost image or created manually. Otherwise, one of the interface instances would require a restart to create the sync file. Until the sync file is created, failover would be controlled via PI server reads.