New: Cookbooks and AI ExplanationsStep-by-Step recipes to solve problems connected to Roadmaps and Cheat Sheets. Need more details? Use AI buttons for structured and simple explanations with concrete examples throughout the whole platform.Take a look
Your last full rollout took the whole API down at once, and everyone found out from the pager, not from a dashboard.
What you'll have at the end
A canary release that starts at five percent of live traffic and keeps widening only while its failure rate stays inside the range the stable version normally runs at, catching or clearing itself with nobody watching a dashboard.
You need
A FastAPI service already running as a Kubernetes Deployment, with Argo Rollouts installed in the cluster, a metrics system already scraping its request counter, and steady enough traffic that a five percent slice is a real sample rather than a rounding error.
Not covered
Adding a second or third gating metric once this single-metric gate has earned the team's trust.
Leans on
Does your rollback restore the last working version?
this recipe's automatic revert only helps if the underlying rollback actually restores a clean state, which is what that recipe proves.
A feature flag that turns off a bad release in seconds
when the broken behavior lives inside one feature rather than the whole release, a flag reverses it in seconds without waiting on this ladder's pause windows.
Checked 20 Aug 2026
Part of the Backend Deployment cookbook