giuseppelt
It's a deep dive, also with the deployment process.
At the end, also some info on how to manage a monorepo with all service systems and components.
Any suggestion or advice is appreciated.
Basically, I think that rolling restarts of the worker instances, just does the trick in the easiest way. Mind that flymachine instances, are pretty quick to boot. While an instance is restarting, others still serve requests so the system is still up.
Build your own FaaS: Self-host workers, lambda, JavaScript functions
It leverages workerd, the Cloudflare worker runtime. And as deploy tutorial, it include a guide to host everything on Fly.
It support container-based deployment, so docker is your friend here.
The complete project is on: https://github.com/giuseppelt/self-workerd
Mind this a proof-of-concept, public exposed systems require more work.
It support container-based deployment, so you can host wherever you like. I published a step by step guide for any one curious: https://www.breakp.dev/blog/build-your-own-faas/
Mind this a proof-of-concept, public exposed systems require more work.
Build your own FaaS: Self-host Cloudflare workers on Fly and everywhere
Architecture of an Early Stage SaaS
Some topics discussed:
- How to design a low-maintenance architecture?
- Which cost-effective providers to use?
- How to deploy to production with ease?
- How to manage a monorepo with all service systems and components?
Any suggestion or advice is appreciated.