SRE · August 2026 · 6 min read
If nobody is on the hook, the build is incomplete
A platform that nobody operates is a project that ended at go-live. Site reliability engineering is the operate motion: what “up” means, who is paged, and how the next change does not repeat the last incident.
Say what “up” is
An SLO is a number a platform owner can live with — availability, latency, freshness — not a slogan. It tells you when to slow a release and when the error budget still has room. Without it, every outage is a surprise and every feature is an argument.
See it, then page it
Observability is not a dashboard wall. It is the signals that tell you the SLO is threatened, with enough context to act. Pages that cannot be closed are noise. Runbooks that exist only in someone’s head disappear on leave.
Incidents are part of the job
On-call has to be real: a path, a severity, a write-up that changes the system. Blameless is useful only if the change actually lands. That is SRE as Vilwave runs it — not a team that watches graphs while someone else owns the night.
If the build did not include this, it was incomplete. Reliability is not a warranty you buy after the cluster is live.