Stripe uses graph search to cut database incident pages

Stripe has automated part of its database incident recovery by turning its global infrastructure into a graph that software can reason over.

InfoQ reports that Stripe’s engineering team combines graph search algorithms with state machines to compute and execute remediation plans for MongoDB shard layouts. Instead of relying only on static runbooks, the system adapts to the current shape of the infrastructure and chooses a recovery path.

The reported operational impact is concrete. Stripe says the approach reduced database-related pager alerts by about 30%, or roughly 200 fewer pages per year, and eliminated an estimated 12 days of unhealthy database time.

This is not a generic AI agent story. It is a useful example of automation built around explicit models of infrastructure, known states, and controlled transitions. For teams running complex data systems, the lesson is that reliability work often improves when recovery logic is encoded clearly enough for machines to execute and humans to audit.