Distributed Queue Engine

A highly resilient, horizontal background processing orchestration layer backed by Redis and BullMQ, designed to offload heavy I/O workloads from the Edge API.

Worker Architecture

System Resiliency Measures

Vertical & Horizontal Scaling

The API servers and Worker processes are explicitly separated into different entry points (api/index.ts vs worker/index.ts). This allows DevOps engineers to scale the API cluster dynamically based on HTTP traffic, while independently scaling the Worker cluster based on Queue depth metrics.

Guaranteed Execution via Redis Hash Tags

Queue keys are strictly defined utilizing {} hash tags (e.g., {emails}:outbound). This guarantees that all internal sorted sets, lists, and metadata keys belonging to a specific BullMQ instance reside on the exact same hash slot within a Redis Cluster pipeline, preventing cross-node CROSSSLOT transaction errors.

Exponential Backoff Safety Nets

Jobs simulating arbitrary external SMTP failures are returned to the queue and delayed utilizing type: 'exponential'. This protects external API limits and auto-heals transient network stutters, ultimately routing chronically failing operations to a Dead Letter Queue (DLQ).