Skip to content

Infrastructure

The underlay, hardware and other resources you need before a platform can be deployed.

This is the provider's part: the hardware and the landscape — storage presenting its resources, a network ready to carry traffic, delegated subnets, and hosts reachable over IPMI. Claugine installs everything else, from green field: the hosts' operating systems, OpenNebula, VyOS and the portal.

Claugine sets no hard restrictions on hardware: the logic of the platform stays the same whatever it runs on. What changes is how effective the platform is.

What an effective platform needs

Resource Recommended
Network Two fabrics: one for data, one for storage. See Underlay.
Storage NFS for VM disks, iSCSI for backups.
Servers Physical servers connected to both fabrics, managed through IPMI.
Public IP subnets Delegated public blocks for projects and published services.
Internet connectivity At each site.
Private inter-site link Optional, but very useful.

Using what you already have

You can start with the equipment you have. The minimums:

Network

  • One network fabric is enough.
  • 10G is the minimum. 1G is technically possible, but we don't work in that environment.

Servers

  • At least two 10G ports.
  • IPMI with the Redfish API.
  • Hardware that is 10 or 15 years old is fine. Use common sense when choosing servers, though: today power consumption matters more than buying cheap, ancient servers.

Storage

  • NFS is what OpenNebula works with best.
  • Any block storage, such as iSCSI or Fibre Channel (FC HBA), is fully acceptable. Claugine adopts it in one of two ways:
    • native Shared LVM, provided by OpenNebula;
    • OCFS2, a cluster shared file system.
  • Hyper-converged setups (HCI) are built with LINBIT.
  • Ceph is never used under the platform. It can serve as secondary storage, or as the underlay for S3 provided by the platform.

Resources as Claugine sees them

flowchart TB
  site["Site<br/>physical location / availability zone"]
  site --> net["Networks<br/>VLANs and their private subnets"]
  site --> pub["IP subnets<br/>delegated public and corporate blocks"]
  site --> sto["Storage<br/>any type, including local disks of servers"]
  site --> srv["Servers<br/>connected to the networks, ready to use the storage"]
  net & pub & sto & srv --> plat["Platforms"]
  • Site: a physical location, an availability zone.
  • Each site has:
    • networks: VLANs and the private IP subnets that go with them;
    • IP subnets: the delegated public blocks (and the corporate ones), recorded on the site's IP subnets tab. Each is carried by one exchange network, and projects get their addresses from them;
    • storage of any type, including the local storage of servers;
    • servers: physical servers, connected to the networks and ready to use the storage.
  • Each site also runs its own core services (not projects'): name servers, admin jump hosts, the log collector, the platforms' front ends, route reflectors, the gateways of the services subnet, backup storage and the portal's own VMs. They are recorded in the portal, deployed either on existing hosts or from scratch as VMs.
  • These resources are then organized into platforms.

How each object is recorded is described in Data model.