Managed infrastructure

Managed Infrastructure is the operator-facing area for capacity that backs GlobalStacks-managed clusters.

It is permission based. The section appears when the active organization has managed_infrastructure_enabled and the user has the required managed infrastructure permissions. The system organization is the first user of this model, but it is not hardcoded as the only possible owner.

Main objects

• Providers: GlobalStacks Internal first, with external vendors registered for future provider extensions.
• Capacity pools: what can be provisioned in a region with a runtime profile.
• Machines: enrolled or planned bare-metal hosts.
• Runtime profiles: approved runtime and isolation targets such as Firecracker micro virtual machine (microVM).
• Regions: allowed placement locations.
• Rate cards: wholesale and retail pricing.
• Settlements: provider accounting records.

Capacity pools

A capacity pool defines what can run:

• provider
• region
• runtime profile
• runtime kind
• isolation kind
• hardware class
• central processing unit (CPU), memory, storage, and sandbox slots
• networking capabilities
• rate card references

It also defines provisioning policy:

• allowed customer segments
• allowed cluster sizes
• allowed runtime profiles
• allowed sandbox image families
• max resources per sandbox
• public ingress policy
• persistent volume policy
• instant, requested, or operator-only provisioning

Machine provisioning

For GlobalStacks-owned hardware, operators can create a managed-capacity enrollment token, copy the installer command, assign the machine to a pool, run readiness checks, run burn-in, and approve activation.

External vendors can register profiles for later provider adapters. Those adapters are not allocation-capable until a future implementation explicitly enables them, validates their bootstrap path, and proves managed-capacity gstacks-agent enrollment.