Overview

A 2019 data migration left 15 million user accounts with missing spending limits records — silently allowing unlimited transactions against BSP regulations. This project designed the remediation system.

Architecture

• Multi-threaded batch processor with cache invalidation strategy for zero-disruption database patching
• Concurrent Kubernetes processing: 3 pods + 3 cron jobs running off-peak within strict time windows to avoid production impact
• Cache invalidation designed to prevent stale limit data from being served during the patching window

Key Contributions

• Architected the end-to-end remediation approach with no service downtime
• Designed execution schedule to work within BSP-mandated time windows
• Coordinated with infrastructure team for Kubernetes pod scaling and monitoring

Tech

Java Spring Boot Kubernetes CronJobs Multi-threading Distributed Caching AWS Alibaba Cloud