[ianceicys]

We keep master to be the version of the code in production. So any pull request that has to be merged to production is code reviewed, automatically deployed to production via spinnaker in a canary, blue-greened across production, and with splunk and new relic monitoring we know if the code in the PR affects production, and if it fails after 15 minutes in production we rollback the code all the way (back in stg\dev too). One of the key things we leverage is feature flagging and ALL code is behind feature flags using launch darkly...so we do controlled exposure of the code (enabling the feature) for specific users\segments of our population until we enable it for all users.

Biggest issue is the delay between code being deployed and enabled in production and users actually using the feature which allows us to get meaningful metrics on the viability of the feature.

A DevOps Definition of Done is working software collecting telemetry against the intended business objectives.

https://docs.microsoft.com/en-us/azure/devops/learn/what-is-...