Independent plants
Every plant runs its own MaestroHub and keeps its data. A separate [Fleet Manager](../architecture.mdx#fleet-manager--fleet-control) deployment holds the licences, the shared single sign-on settings and the health of every instance.
- Scenario
- 4 of 5
- Shape
- Tier 1 or 2 per plant · Fleet Manager
- Applies to
- 3.0
Choose it when plants must keep running with no WAN, data residency rules apply per country, or plants differ in size.
- MaestroHub
- Reads machines
- Data
- Buffered on disk
- Management
- Your tooling
What you install
| Level | Component | How it runs | Recommended hardware |
|---|---|---|---|
| L5 | Fleet Manager | Its own binary or Helm chart | A small server or container host |
| L3 | MaestroHub in a small plant, Tier 1 | One binary as a native service | 8 vCPU, 16 GB RAM, 200 GB SSD |
| L3.5 | MaestroHub in a large plant, Tier 2 | The single-node chart on K3s | 16 vCPU, 24 GB RAM, 500 GB SSD |
Each instance sends a heartbeat to Fleet Manager every 5 seconds over HTTPS and gets its licence back in the reply. Fleet Manager shows each instance's version and health, sends alerts, and distributes single sign-on settings (OIDC, SAML, LDAP). Every connection is opened by the instance, so no inbound port is needed at a plant.
When something fails
| What fails | What happens |
|---|---|
| The WAN | Nothing changes inside the plants. Heartbeats resume when the link returns. |
| Fleet Manager, for longer than the grace period | Each instance keeps its last licence for 7 days by default, then falls back to its local licence file (or the evaluation without one) and keeps retrying. |
Limits. Each plant has its own namespace; there is no view across plants. Upgrades are applied per deployment: Fleet Manager does not roll out software versions.
Previous: One hub, many sites · Next: Three-layer namespace · How to read the drawings: Deployment scenarios