Network & Infrastructure / High Availability & Recovery Design
N+1 capacity under actual failure placement
A cluster has one spare node on paper but overloads when its largest node fails. Capacity counted by node quantity can hide unequal sizes and workload-placement constraints.
Environment and inputs: Routed and clustered services; define failure ownership, fencing and application recovery requirements before testing. Diagnostic/design example; not tested on production equipment.
Collect the evidence
Model CPU, memory, network, storage and placement requirements after each relevant failure. Include reservations, licensing and affinity rules that prevent workloads from using nominally free resources.
Explain the result
N+1 is meaningful only relative to the failure unit and usable surviving capacity. An average node size does not prove the largest or most constrained workload can be recovered.
Plan the fix
Size reserve against the tested worst required failure and enforce placement consistent with that reserve. Keep recovery traffic and cold-start demand in the degraded-mode budget.
Check the result and rollback
Exercise representative evacuation and peak demand while measuring service objectives. Revert a consolidation change if the surviving system cannot host required workloads without violating capacity or placement constraints.