I spent a day or two tuning our release pipeline for our
@honojs API on
@CloudflareDev workers.
Our product is in the critical path of our users, so we really can't accept high latency, regardless of where our user is, but measuring that as part of the CI/CD pipeline is not something I've seen before, here's a quick rundown of how we do it:
1. run unit tests locally
2. deploy to preview and run integration tests
3. deploy to canary (100% identical to our prod setup, except hostname)
4. run integration tests against canary
5. run global latency benchmarks against canary (we relay requests through
@vercel to measure latency from 18 different locations.
7. deploy to prod with confidence
As always, everything is 100% open source