Fly.io has resolved a major incident affecting Upstash Redis services in the fra and gig regions. From approximately 12:00 to 13:05 UTC, some clients experienced connection timeouts to their Redis databases. The outage affected databases hosted in those regions, as well as clients whose connections were routed through them, even if their database was located elsewhere.
Fly.io is a cloud infrastructure provider offering container deployment and managed services to developers and teams building web applications and backend systems.
During a planned rolling upgrade, replicas in fra and gig failed to complete the connection-draining process before being returned to service. The incompletely drained replicas were unable to accept incoming connections, resulting in timeouts for affected clients. Fly.io identified and resolved the issue, then implemented monitoring to verify the fix.
The company announced plans to add tooling for safe replica restoration after maintenance, introduce additional safety checks to the upgrade process, and improve the overall resilience of upgrade procedures to prevent similar incidents.
