// the find
Mikaelemmmm/go-zero-looklook
🔥基于go-zero(go zero) 微服务全技术栈开发最佳实践项目。Develop best practice projects based on the full technology stack of go zero (go zero) microservices.
A reference implementation of a microservices system built on go-zero, using a fake homestay-booking app as the domain. It wires together the entire go-zero ecosystem: API+RPC layers, Kafka via go-queue, asynq for delayed/scheduled jobs, DTM for distributed transactions, Jaeger tracing, Prometheus+Grafana monitoring, and a full CI/CD pipeline. Aimed at Go developers who want to see how all these pieces connect rather than learn any single piece in isolation.
- The API-as-aggregation, RPC-as-business-logic split is applied consistently across every service — this is the correct go-zero pattern and seeing it repeated across order/payment/travel/usercenter makes the convention stick.
- Using go-stash instead of Logstash is the right call; Logstash's JVM memory footprint on a small cluster is genuinely painful and this project shows a working alternative without ceremony.
- 15-chapter Chinese documentation covers the full lifecycle from local docker-compose dev to k8s production deployment — more practical depth than most example repos at this scale.
- asynq for delayed queues and timed tasks is a good pragmatic choice over a full job-scheduling system; the project shows it sitting alongside Kafka publish/subscribe rather than trying to replace it.
- Last pushed January 2025 and go-zero has kept moving — goctl-generated code templates and the pb generation scripts are likely out of sync with current toolchain versions; you will hit breakage before you run a single service.
- The Jenkins/GitLab/Harbor/k8s CI/CD pipeline is described in prose but the actual Jenkinsfiles and k8s manifests are absent from the repo, so the deployment story is documentation theater.
- The business logic is thin scaffolding: error handling is surface-level, there's no meaningful idempotency on the payment or order flows, and the DTM integration is 'prepared for future use' rather than actually wired in.
- English README and docs are machine-translated with enough ambiguity that non-Chinese readers will routinely have to guess at intent — the project is functionally Chinese-first.