Understand the antalos stack
antalos combines several systems with different owners. Start with the repository map, then follow the machines, network, and data before reading an individual application’s design.
Read the architecture in order
| Guide | What you will understand |
|---|---|
| Repository and ownership | Which directory changes machines, Kubernetes resources, VM workloads, or this website |
| Hosts and Talos | How the physical hosts, control plane, workers, and quorum placement fit together |
| Traffic, DNS, and TLS | How the Azure edge, home HAProxy pair, MetalLB, Traefik, and application listeners connect |
| Storage and data | Where databases, files, objects, and recoverable SQLite state live |
| Debian application VMs | Which features run outside Kubernetes and how Ansible owns them |
| Availability and failure domains | What replicas protect and which shared dependencies can still interrupt a service |
Read an application’s design
Use the Architecture link in the service directory. Each service guide describes its components, state, and failure behavior. The neighboring Use and Operate guides explain user impact and administrator actions.
For a first example, Nextcloud shows a web application with PostgreSQL, Redis, object storage, shared files, and VM companions. Obsidian shows why two database processes do not necessarily form a synchronous cluster. Uptime Kuma shows the difference between a restartable singleton and a ready replica.
Terms used throughout the guides
| Term | Meaning here |
|---|---|
| Desired state | The configuration declared in Git or an infrastructure project |
| Reconciliation | A controller working to make running resources match desired state |
| Replica | Another process or data copy; the guide states which |
| Quorum | Enough voting members to make a coordinated decision |
| Failure domain | A component whose loss can remove several resources together, such as a physical host |
| HA | High availability for a specified failure, with stated dependencies |
| Recovery | Bringing a service back through restart, repair, or restore after interruption |
A workload’s replica count is one part of its design. Storage, routing, identity, controller access, and physical placement determine whether the user can still complete a task.