How Stripe uses graph search and state machines to auto-remediate a global database fleet

This post details how Stripe's automated database fleet remediation system, which models MongoDB infrastructure as a traversable graph and uses pathfinding algorithms for recovery, reduced pager volume by 30% (~200 pages/year) and eliminated 12 days of unhealthy shard states annually. It also highlights the system's ability to support new shard layouts with zero manual effort.