Sentry migrated their issue alerting system from multiple fragmented implementations to a unified workflow engine. The migration involved converting 2 million JSON blob-based alert configurations into 20 million normalized database rows, rewriting condition processing logic to handle both fast and slow evaluations, and running both systems in parallel before switching over. The team encountered a Redis performance incident during rollout due to shared key contention, which they resolved by sharding keys and provisioning a dedicated cluster. The new system now handles all error alerts with feature flags ready for other issue types.
Sort: