Write the operating plan
A service needs more than hardware. Someone must know when to admit work, when to stop, and what evidence allows recovery.
Give each state a release condition
- Installed → accepted
Commissioning evidence
- Available → running
Authorized allocation
- Suspect → quarantined
Approved containment
- Validated → available
Explicit return-to-service decision
Make states actionable
Installed, accepted, available, running, draining, and quarantined describe different conditions.
A failed health check should lead to the defined operational response, not just a different dashboard color. Specify who can change each state and why.
Connect signals to people
For a latency failure, device fault, missing measurement, or cooling alarm, name the responder and the affected equipment.
Changes need compatibility checks, a limited test, stop conditions, and a recovery path. Physical work still follows product procedures and qualified ownership.
Your deliverable
Write a one-page plan for the fictional service from Sessions 41–42. Include service targets, states, monitoring, maintenance, recovery checks, and owners.
Check your answer
If a leak page reaches facilities without a rack identifier and compute operations know nothing, the response is incomplete. Another responder should be able to use your plan without asking the original designer to explain it.