gh0stcloud from the inside

What happens when an app goes live with us?

From the data centre down to a single app, layer by layer. Every level explained so that it makes sense without a technical background, and, if wanted, in as much depth as a platform engineer would ask for.

A model in time-lapse. The parts and the processes are real; amounts, paths and pace are simplified. No live data.

1 · Data centre

It all starts with hardware in Germany.

gh0stcloud runs in data centres operated by Hetzner, a German provider. Hetzner supplies power, cooling, physical access and connectivity. Everything that runs on the servers, we operate.

2 · Servers

Part of a machine, or all of it.

A cloud server is a separated share of a large computer. A layer beneath, the hypervisor, keeps the shares apart. A dedicated server is a whole machine for a single owner.

3 · Environments

Three environments, one job each.

The vault keeps keys, passwords and sign-in. The production environment carries the customers' apps. Every change runs in the test environment before it reaches production.

4 · Customers

Many customers, one cluster, clear boundaries.

Every customer has an organisation. It holds projects, the projects hold namespaces, the namespaces hold the apps. The boundary that matters is the namespace: what runs in it cannot be reached from other namespaces until a connection is approved.

5 · Network

Only published routes lead in.

Traffic from outside comes in through one gate: the platform's entrance. There it goes to the app that published exactly that address, encrypted with a certificate. Everything else finds no open door.

6 · Services

Ready-made apps that work together.

We run Managed Apps for businesses: setup, backups, updates, monitoring. Apps that need each other come as a bundle, such as a mail server with webmail. The apps run under the customer's own domain.

7 · Operations

Described, checked, watched.

What the platform should look like is written in Git. Reconciliation keeps checking that everything runs as described and puts back whatever drifts. Anything new that wants into the cluster has to pass a check first.

8 · Backups and service

Backups, restores, commitments.

We back up the data of Managed Apps regularly into storage outside the cluster, at Hetzner in Germany. If something is lost, we bring it back from a backup.

Managed Basic

Daily backups, kept for 14 days. First response to critical incidents within 4 business hours.

Managed Business

Backups every 6 hours, continuous for PostgreSQL, kept for 30 days. First response to critical incidents within 2 business hours.

Self-operated

An answer usually within 24 hours on business days. Backups are the customer's.

9 · Sovereignty

Customer data stays in the EU, with EU companies.

The platform runs at Hetzner in Germany and is operated by gh0stservice GmbH in Cologne. No group outside the EU stands behind us, and we use no service provider outside the EU for customer data. The platform does not run on AWS, Azure or Google Cloud.