Chapter 1 of 7 · 4 checks · 2 min

Ownership and documentation

Someone has to own the service, and the people paged for it need enough written down to act without its author.

Show

A named team owns the service and is reachableMust
Why it matters and how to check
Every production service needs one owning team, not an individual and not "everyone". Ownership means they fix it when it breaks, keep its dependencies current and decide its priorities.
To verify, find the owner in your service catalog, repository metadata or CODEOWNERS file in under a minute, and confirm that team's on-call rotation includes the service. If you have to ask around to find who owns it, this check fails.
There is a short architecture overview with all dependenciesMust
Why it matters and how to check
The SRE launch checklist starts with the architecture: the components, the request types and what the service depends on. Responders need this to reason about failures they have not seen before.
Write one page with a diagram: the components, the data stores, every upstream and downstream dependency (including third-party APIs), and which are critical versus optional. To verify, ask someone outside the team to read it and name what happens if the database or the payment provider is down. If they cannot, the document is missing something.
A runbook covers the common failures and first stepsMust
Why it matters and how to check
The person paged may not know the service. A runbook gives them first steps: how to check health, how to see recent deploys, how to roll back, how to scale, how to drain traffic and who to escalate to.
To verify, pick two alerts for the service and follow their runbook links. Each should open a page with concrete commands or dashboard links, not a stub. Have someone who did not write it walk through it in staging.
Do
Link the runbook from every alert
Keep rollback and scale-out commands copy-pasteable
Each critical dependency has a known owner and escalation pathShould
Why it matters and how to check
When a dependency fails, you need to know who to contact and how: another team's on-call, a vendor's support line, a status page to watch.
To verify, for every critical dependency in the architecture overview, check that the runbook lists the owner, the escalation channel and any status page. Test one escalation path before launch so the contact details are known to work.