Skip to content

Check: controller validation and responsibility boundaries

Knowledge check · self-test

Check: controller validation and responsibility boundaries

Check request validation, responsibility boundaries, local fallback and stale-topology handling.

Back to the chapter →

Check request validation, responsibility boundaries, local fallback and stale-topology handling.

Write one change record. Name the requester, allowed goal, state time, rule owner, affected switches, safe default, deadline, reply, final service check, and rollback point. Keep the request, controller decision, device action, and user result separate.

Test stale state, two requests that conflict, one switch that rejects the rule, a lost controller, a late reply, and a rule that installs but breaks the service. Compare controller records with switch state and the receiving service. An accepted request is not proof of useful forwarding.

Use a short release check. Request one allowed change and one forbidden change. Delay one switch reply and remove the controller before the next request. Staff should see the rule state, service result, owner, and safe fallback without guessing.

Repeat the check after a policy, switch, link, controller service, or state source changes. Save the failed case because it shows which proof or rollback step needs work.

Close with four plain questions. Who asked? What did the network change? Did the real service still work? What safe rule stayed when the controller went away? If staff cannot answer all four from one record, the change path is not ready.

Start simple: name one service request and the controller responsibility it creates. Then the controller architecture can be judged by state ownership, interface clarity, rule installation, and failure behavior.

Back to SDN Controllers: Architecture Foundations: the section this check follows