Surviving failures.
WHAT CAN FAIL
A cable A switch A switch port An upstream provider A fibre route Power to network equipment
WHAT REDUNDANCY REQUIRES
Two of each, on genuinely separate paths.
WHAT DEFEATS IT
Two cables in the same tray Two connections to the same switch Two providers using the same physical route Both switches on the same power feed
THAT LAST ONE
Commonly overlooked, and it makes switch redundancy meaningless.
WHAT PROTOCOLS PROVIDE
Automatic failover when a path fails.
WHAT TO VERIFY
That failover actually works, by testing it.
WHY TEST
Configurations that should fail over frequently do not, and nobody discovers this until it matters.
WHAT TO TEST
Removing one path deliberately, during a maintenance window.
WHAT TO MONITOR
Link status Whether you are running on a single path Error rates on each link
THAT SECOND POINT
Running on one path after a silent failure is a common and dangerous state.
WHAT TO ALERT ON
Loss of redundancy, not only loss of service.
