Failover Scenarios

Loss of the HA node’s management interface will prevent automated failover due to inability to communicate with the witness server. However, this will not directly impact the data availability. The system administrator would still be able to failover any Resource Groups using the second, healthy Head Node.

To overcome this edge case, an additional management interface can be established over the data aggregate and designated for HA communications.

Loss of data network entirely is highly unlikely when it is configured as an aggregate of two or more physical ports. In this unlikely event or when the Head Node is configured with only a single data interface, HA can be configured to failover on data interface loss.

RMM and iLO communication for HA functionality is only used as a third means of power state verification. Loss of lights-out interface has no impact on system functionality given the management and heartbeat interfaces are healthy. If all means of communication are unavailable for a given node, a failover event will take place.

Manual failover is an action triggered by the system administrator to initiate a Resource Group(s) migration to the adjacent HA node. This action is typically performed as part of the solution maintenance or during upgrades.

Automatic failover takes place during an HA node failure. In this event, the surviving node will take over all the Resource Groups.

After automatic failover takes place, it is advised to disable the failed node to prevent further action. For example: Resource Groups can be configured to have a preferred node. Disabling a failed node will prevent resources from failing over back and forth in the event this HA node is exibiting an inconsistent behavior.

Once the issue is cleared up or the failed node is repaired, it can be re-enabled and returned back to service.