Bring your own cloud
PicoPaaS is heading to bring your own cloud, so you can run it where you already have your data and your bill. The first step shipped: a CLOUD_PROVIDER setting, DB flavors and tighten-only isolation ranges. Nothing changes for users.
PicoPaaS is heading to "bring your own cloud". Chas, who owns the project, has set the direction: the platform should run wherever you or your company decide to deploy it. Today it runs on DigitalOcean, and picopaas.xyz still does. Nothing changed for users. But we just shipped the first step, and we wanted to say where this is going.
Why
If you already have a cloud, you should be able to run PicoPaaS in it. That way you keep your data, your bill and your network. Platform work continues alongside Kaisen.
What shipped
- One
CLOUD_PROVIDERsetting. Every DigitalOcean-specific piece now sits behind an interface: app buckets and keys, the backup target, DB host metrics and the storage check. DigitalOcean is the only implementation for now, and it behaves exactly as before. - A per-database-server "flavor", so differences between managed databases live in one place instead of being scattered through the code.
- Tenant isolation ranges now come from config. They can only ever be tightened. An operator can add ranges, but the defaults are always excluded, so a bad setting cannot open anything up.
How we checked
We ran a full regression against production: e2e, quota flips, backups and downloads, rotation, rollback, volumes, and the isolation pentest suite. We also had Kaisen build a guestbook on PostgreSQL, then changed it and restarted it. The data survived both. That build cost under two US cents.
What's next
First a "generic" target: any Kubernetes cluster, with Garage for S3-compatible storage, plus PostgreSQL and MySQL, tested in CI. Then Helm packaging and external-dns. After that, AWS and GCP. Which one comes first depends on where we can get test credits, and we are not promising dates.
Azure comes later. Its object storage has no S3 API, and apps keep the same S3 env vars everywhere, so it needs more work than the others.