INC-2026-508: Notification Service Sev2 Incident
Date: 2025-07-15 Service: Notification Service Severity: Sev2 Incident Commander: Morgan Lee Jira References: ACME-5015, ACME-5016 GitHub References: GH-5015, GH-5016 API References: API-NOTIFICATION-01
Summary
The Notification Service experienced customer-visible degradation caused by stale cache entries surviving a blue-green release. The incident affected dependent paths that include Authentication Service, Checkout Service, Billing Service, Inventory Service, and Notification Service, with the strongest impact localized to Notification Service.
Timeline
- Detection: synthetic checks and service SLO burn alerts fired.
- Triage: on-call engineers reviewed traces, dependency dashboards, and recent deployments.
- Mitigation: traffic was shifted, retries were reduced, and the affected worker pool was drained.
- Recovery: error rate returned to baseline after remediation was deployed and queues caught up.
Root Cause
The proximate cause was stale cache entries surviving a blue-green release. The deeper contributing factor was incomplete validation of the Notification Service operational contract against API API-NOTIFICATION-01 and insufficient pre-release checks for high-cardinality traffic.
Customer Impact
Customers saw intermittent failures, elevated latency, or delayed asynchronous updates depending on where their request crossed the Notification Service boundary. No permanent data loss was identified.
Remediation
- Track remediation in Jira tickets ACME-5015, ACME-5016.
- Implement code and test changes in GitHub items GH-5015, GH-5016.
- Update the relevant runbook for Notification Service and replay this scenario in staging.