Durable queue infrastructure · Go

RivetQ

Workflow-heavy systems need durable job state, lease-based delivery, and retry semantics, not a Redis list pretending to be a queue.

View source

What I built

A Go service with a segmented write-ahead log (64MB segments, CRC, replay/compaction) and HashiCorp Raft for HA. Jobs move through Ready → Inflight (visibility timeout) → Completed / DLQ with exponential backoff + jitter. gRPC + REST APIs, a Next.js admin dashboard, and Prometheus metrics.

Stack

GoWALRaftgRPCNext.js

Status

Operational core (~6K LOC Go); admin surface + Prometheus shipped.