Comparison
Automated rollbackvsProgressive delivery
Automated rollback
error rate crossed the threshold ninety seconds in and the system reverted to the previous version without waking anyone.
Reverting a release automatically when post-deploy signals breach a threshold. It converts a class of incident into a non-event, and it depends entirely on the previous artefact still being deployable — which is a statement about migrations, not about the pipeline. Forward-only changes to schemas and data are exactly what make automated rollback impossible, which is why they deserve the scrutiny they get.
Full entry →Progressive delivery
the release goes to one percent, then five, then twenty-five, and each step is only taken if the metrics stayed flat.
Releasing to progressively larger portions of traffic with automated analysis deciding whether to continue, pause or revert. It is canarying with the judgement automated and formalised into steps. It requires two things most teams lack before they start: routing that can split traffic by percentage, and a metric trustworthy enough to abort a release on.
Full entry →