ForkTex Cloud — Declarative Deployment

Romania

Our own deployment platform: one manifest describes a system, and the platform provisions it, routes it, secures it, and cuts traffic over without downtime.

Client
ForkTex
Industry
Cloud Infrastructure
Period
Jun 2024 — Ongoing

Outcomes

8

Production stacks deployed through it

Deploys
Blue-green over Docker
Infrastructure
Hetzner, European

Technologies

The situation

Every project needs somewhere to run, with TLS, a database, backups and a way to deploy without downtime. Renting that from a managed platform is the obvious answer and was ours for a while.

What we built

The manifest

A forktex.json describes services, the infrastructure they need, and the deployment strategy. Each service carries a type — compute, persistence or daemon — and the type is what the platform reasons about. The proxy layer is generated from the service list rather than written.

Provisioning

Hetzner servers, Cloudflare DNS and Let’s Encrypt certificates, provisioned by Ansible. Secrets are referenced rather than embedded and resolved at deploy time, so the manifest is safe to commit.

Deployment

Blue-green over Docker. The new slot passes health checks before the proxy moves, then the old slot drains for a grace period and stops.

What mattered

The property that matters is not zero downtime but that rollback is a proxy change. When reverting is a rebuild, teams deploy less often in bigger batches, which makes each deploy riskier.

Outcome

Eight production stacks deploy through it, including this website. We carry the backups, patching and provisioning a managed platform would absorb; that cost is only repaid because the same platform runs every project.