Building a Deployment Checklist That Actually Prevents Production Incidents

A deployment checklist should focus on application risk rather than infrastructure items. Application risk includes factors like bad configuration, unsafe schema changes, and unready dependencies that can cause user harm. Infrastructure risks, such as valid certificates and correct autoscaling rules, should be handled automatically by the platform.

To create an effective checklist, teams should identify specific incidents from the past and determine what checks could have prevented them. This involves automating tasks that require human judgment, as humans are prone to error when fatigued. Some key application risks to consider include the ability to undo changes quickly, ensuring necessary configurations are in place, validating the safety of database changes, ensuring honest health reporting from the app, and gradually rolling out changes to users to minimize the impact of mistakes.

Database migrations pose unique challenges for rollbacks, as they cannot be undone as easily as code changes. By focusing on these application risks and automating as much as possible, teams can create a deployment checklist that truly prevents production incidents.