Data model¶
What the portal works with, who defines it and where it lives: the infrastructure records, project resources and how they are stored in OpenNebula. The portal says project everywhere; in the code the model is still called Tenant, so the record names below keep it where they name a model.
1. Sources of data¶
| Data | Defined by | Lives in |
|---|---|---|
| Infrastructure | resource managers and support engineers, in the portal | portal records (sites, IP subnets, profiles, hosts, storage, networks, platforms, clusters, core services). entered and edited only in the portal |
| Secrets | the same people | portal records, encrypted (Infrastructure → Secrets); records refer to them by id |
| Organizations, projects, orders, resource records, addresses | customers and staff, through the portal | the portal database (desired state) |
| What OpenNebula holds | the platforms | read-only by the Platforms timer (actual state) |
2. Principles¶
- One place to edit. Every infrastructure object is a record edited in a dialog of its list; no file is read at run time. A record that others refer to (a volume under a datastore, a network under a platform network, a host in a cluster, an IP subnet in use) cannot be deleted.
- Names are unique where OpenNebula needs it. Objects created in OpenNebula (clusters, datastores, networks) carry the site in their name,
payload1-vms01_dc1, so federated zones never clash. - Sizes are in GiB.
- States: planned (not installed), active, retired (removed, kept until taken down). Platforms are planned, deployed or retired.
- Desired vs actual. The database is the desired state; project records are versioned (append-only); projects are archived, never removed from the database.
3. Infrastructure records¶
server profile ─┐
site ───────────┼── host ───────────────┐
│ │ │
├── storage (nfs / iscsi / local) ── volume ──┐
├── IP subnet (inet / corpnet) ─┐ │
├── network (core-… / core / ix / evpn / nfs / iscsi / public)
│ ├── ranges │
│ ├── IP subnets it carries (ix) │
│ └── BGP peer groups (rrs, igw, egw) ─ secrets
├── core service ── nodes (hosts or VMs) │
└── platform (kind, front end, mode) │
├── datastore (vms / images / files / backups) ─ volume
├── network (vlan / bridge) ─ site network + ranges
└── cluster (kind) ─ hosts, datastores, networks
4. Site resources¶
Site¶
Code (name), site code (number, fixed once set, part of the AS numbers), country, city, time zone, NTP (up to 4), DNS (up to 3), green-field DNS (up to 2), IPMI settings, management hosts and ports, jump hosts, log collectors.
IP subnet¶
The public and corporate address ranges of a site, defined on the site's IP subnets tab. Projects get their addresses from them.
| Field | Meaning |
|---|---|
| subnet | network address and prefix, e.g. 203.0.113.0/24 |
| max | the largest block (smallest prefix length) a project may be given from it, enforced when addresses are handed out; empty means the subnet itself |
| routing | BGP (ix): announced over BGP through an exchange network; static gateway: routed by a gateway inside the subnet (required then); organization: the organization routes it, the portal only delegates blocks |
| type | internet or corporate network, detected from the address: a private range is corporate, a public one internet |
| network | the exchange (ix) network that carries it, chosen on the Networks tab |
- A subnet cannot be edited after it is added. It cannot be deleted while it is in use: carried by a network, or with blocks assigned inside it.
- An ix network carries subnets of one type only (its own: internet or corporate), and can carry several. A subnet belongs to one ix network only.
- The prefix lengths offered when a project is ordered are /31, then /29, /28 … down to the site's most permissive max for that type.
- The subnet's tab also shows its use (assigned addresses of the total), its assignments, overlapping assignments and assignments outside every subnet.
Server profile¶
| Field | Meaning |
|---|---|
| chassis | vendor, model, IPMI type |
| CPU | vendor, model, sockets, cores per socket, MHz — physical cores feed the capacity view |
| RAM | total GB, DIMM details |
| boot | boot device and its volumes (name, size, file system, mount path) |
| NICs | cards: vendor, model, type (media), speed, number of ports and their OS names; the ports behind the aliases nic-data1/2 and nic-storage1/2 and the bonds they make (one or two fabrics) |
| GPU | model, count, VRAM |
| reference host, check | the host the profile was made from; the last consistency check of every host of the profile |
| boot storage | how the boot device is split when a host is deployed (see Boot storage below); it must fit the profile's boot device; a profile made from a host gets the first that fits |
| provider's part | once the hardware is verified: the BIOS and iLO firmware versions a host must have to be deployed, the BIOS settings it gets (one "Attribute = Value" per line; PXE off on all NICs and the site's time zone are always set, read-only), the iLO services left on (ssh, IPMI over LAN) |
Server profiles are listed under Infrastructure → Hardware profiles → Servers.
Boot storage¶
Infrastructure → Hardware profiles → Boot storage: how a host's boot device is split when it is deployed. Always GPT with EFI (512 MiB) and /boot (1 GiB), then an LVM volume group with the volumes listed: mount point, fixed size in GiB or the rest, file system; a swap volume replaces the swap file. A mount point is written one way, and /boot or anything under it is refused. The installation brings one, standard: / 50 GiB, /var/lib 200 GiB, /var/log 100 GiB, /home the rest. A boot storage in use by a profile is not deleted.
Host¶
Name (the hostname), IP for ssh, IPMI address, type and https port, IPMI user (a secret), state, server profile (optional: detected from the hardware when added; a new profile is made when none matches), hardware facts as reported by the OS and IPMI, health as read from IPMI.
Storage and volumes¶
| Type | Storage | Volume |
|---|---|---|
| NFS | one target address, fstab parameters | name, export, size GiB, clients |
| iSCSI | several target addresses, port, fstab parameters | device (as in lsblk), size GiB, initiator IQNs |
| local | the host | device (as in lsblk), size GiB |
Network¶
Name, type, interface, VLAN, MTU, subnet, gateway, state; ranges (name, start, length); BGP: local AS, the IP subnets it carries (ix: announced to the remote AS), remote AS (ix), and the peer groups rrs (addresses), igw and egw (addresses and subnets), each with the range its hosts came from and a password set from Secrets. By default an exchange network gets the ranges ext-gws (.0/30) and rrs (.241–.243: the three route reflectors), an EVPN (VXLAN) network the range rrs (.241–.243). The routers name the route reflectors one by one, as the network's list of route reflectors: the egw, cgw and the route reflectors' own configurations have no listen range for them. When a router is updated, what an older configuration had for them (a listen range, the group of the older .240/30 range and the rules that use it, the neighbor at .240) is removed; a peer or a rule of its own that is not part of that configuration stays. A route reflector is configured with its own address on the network, as its definition gives it. When that address is no longer one of the network's route reflectors, its configuration is incomplete and a redeploy is refused, even a forced one: correct the address in the service's definition first.
| Type | What it is |
|---|---|
| core-services, core-cloud, core-vyos, core-s2s, core-kvm | one of each per site: the core services' public addresses, the core servers' VMs, VyOS management, the L2 channels' transport, the KVM hosts |
| core-lm | one per site: live migration between hosts; a deployed host gets its VLAN and bridge (one-lm) with the host's own number in the network, as on the core network. Today the VLAN is created and not yet used for migrations |
| core | any other core network |
| ix | an exchange network to the internet or to the corporate network |
| evpn | the VXLAN/EVPN underlay |
| nfs, iscsi | storage |
| public | a public overlay of an ext-managed platform |
5. Platforms¶
| Record | Fields |
|---|---|
| Platform | name, kind (mgmt, shared, ext-managed: managed by Claugine, its payload by a third-party system), state (planned, deployed, retired), provider, super admins (secret), front end: mode (single, local, primary, secondary → primary), runs on (VMs or the host OS), zone name, the VIP with the front end's host name, exactly three nodes (each with its own name and IP, other than the VIP; on a deployed platform only the nodes' IPs change; an empty IP is taken from the last refresh, matching the node's name), OneGate through hosts (or the VIP), OpenNebula credentials (secret); the public FQDN of a shared platform (its front end on 443: FireEdge, XML-RPC, OneFlow); operations mode (read-only, or execution: the portal provisions projects) and network function image (default VyOS Router 2026.03); the platform's own VMs (group oneadmin) as read with the core services |
| Datastore | name, type (vms, images, files, backups), the storage volume, path, the share kept free (10 % by default, 20 % for backups); backups: where restic keeps them (an iSCSI volume behind the restic VM, or an S3 bucket) and its password |
| Network | name, type (vlan: an 802.1Q network on a site network; bridge: an existing bridge), the site network, the ranges to create |
| Cluster | name, kind (mgmt, shared, dedicated, platform, edge, ext-managed), state, hosts, datastores and networks of the platform, vCores per pCore (a whole number; 1 with GPUs, else 10), spare hosts (N+X; 0 with GPUs, else 1), reserved CPU and RAM (5 % each), the bundle size of a GPU cluster (with an optional storage per bundle). The quota code is not shown. |
| Core service | site, service type (bgp-rr, cloud-egw, cloud-igw, cloud-gw, portal-gw, ns, logs, admin, backup, core-vm), name, description, deployment (on hosts, or VMs from scratch), state (planned, active, retired), platform hosting the VMs, platform served (empty: core), template, differences from the template, image, ssh port, secret for ssh, secret for the service, ufw rules, checks; from scratch: vCPU, RAM, disk, NIC 1 network and gateway, further NICs (network, gateway), routes (NIC, subnet, next hop, metric); when it was read. |
| Core service node | the service, name (the VM's on the platform or the host's), host name, an address per NIC, the address and ssh port the portal logs in at, secret, the host (on hosts), the login password hash it keeps when redeployed (encrypted), what it reported when read and the last error. |
6. Secrets¶
Every secret is a record of one type, with a name unique within its type, a purpose and a state (active, replaced, archived). Records refer to it by id, so its name can change. Types: Users (user name, password, ssh key), BGP (the passwords of one exchange or EVPN network of a site: rrs, igw, and egw for exchange networks), OpenNebula (a front end: the OS user's password and key, its database name, user and password, OpenNebula's oneadmin password), Keys (public keys), Passwords, S3 (access key and secret key). What is not secret (a database name, a public key, known hosts) is kept beside the encrypted values. New passwords follow the password policy. Archived secrets are kept for the record; values are shown only on purpose and every view is logged.
7. Checks¶
- a secondary front end has a primary; a host belongs to one site and its clusters to one platform;
- object names created in OpenNebula end with
_<site>; - shared front ends are separate: no exchange network and no host may be used by two of them;
- BGP peer hosts: route reflectors addresses only, igw and egw addresses or subnets, all inside the network;
- an ix network carries only IP subnets of its type, and an IP subnet is carried by one ix network only;
- Check on a server profile compares the hardware of every host with the profile.
8. Project resource records¶
A project has one record per cluster (quota code) on each site where it has resources; project + site + quota code is unique. Records are append-only versions with an effective date.
| Field | Level | Meaning |
|---|---|---|
| cpu, ram_gb, gpu | cluster | vCPU, RAM GB, GPUs (in bundles: one GPU with a fixed share of vCPU and RAM) |
| disk_gb, images_gb, files_gb | cluster | storage: one number for VM disks, images and files and the image and file parts of it, see storage quotas |
| backups_gb | site | backup datastore space: one figure per site, shown on the site's first resource pool; the platform gets it once |
| connection types | site | routed connection inet or corpnet (never both), none when only L2; see the network page |
| igw_type, inet_mbps, inet_main_prefix, inet_public_prefixes | site | internet: the project gateway's type (vyos advanced (VyOS), onevr native (Service Virtual Router), custom the project's own VM), bandwidth, main and additional prefix lengths |
| corpnet_prefixes | site | corporate network prefix lengths |
| l2_vlans, l2_vnets | site | external L2 channels: VLANs (entered by support) and the project's networks they connect to |
| date_created | cluster | first day the resources are available |
| date_review | — | the day the current status (or the POC) ends and is reviewed; database only |
9. Storage quotas¶
- One storage number covers VM disks, images and files: a resource pool's storage is images + files + VM disks. Images and files are set for each pool; VM disks are the rest, on the pool's cluster. Project admins split it on the project page without approval. A new resource pool has no images and no files: all of its storage is VM disks until the project admins split it on the project page.
- On the platform, a pool's images quota goes to the image datastore of its cluster; pools whose clusters share an image datastore share its quota (their sum). The platform's other image datastores are closed to the project (0).
- Files: one file datastore per site, shared by all the project's pools there (their sum).
- Backups is separate from storage and is one figure per site, shown on the site's first pool, 0 by default; the platform gets it once.
- These are the rules the deployment applies and the consistency check compares (Site → Platforms, Consistency with the portal).
- An archived project has nothing left on the platforms: its record stays in the portal as history.
10. Project status¶
A project has a type (cloud or corporate), a stage (POC or client; every project starts as a POC), an approval (waiting, approved, rejected, cancelled) and a status. The status table gives what each status means for the project:
| Status | Account | VMs | Connectivity | Resources |
|---|---|---|---|---|
| ok | active | running | working | kept |
| paused | blocked in OpenNebula (portal login works) | running | blocked | kept |
| suspended | blocked | powered off | blocked | kept |
| saved | deleted | deleted | — | one full backup of every VM is kept |
| archived | deleted, portal login blocked | deleted | — | nothing left on the platform; the records stay |
| being deleted | the portal is removing the project from the platform, then it is archived |
ok ──► paused ──(2 weeks)──► suspended ──(1 month)──► saved ──(1 month)──► archived
▲ │ │
└─────────┴──────────────────────┘
Statuses are changed by hand; the only automatic step is saved → archived, a month after saving. date_review is the day the current status ends (paused + 2 weeks, suspended + 1 month, saved + 1 month); until then a paused or suspended project can return to ok. A POC is reviewed a month after it is created. archived is final. The database row itself is never deleted. Archiving a project, by hand or automatically a month after saving, releases everything it still has, the same as a deletion: on the platforms and in the portal (quotas, networks, subnets, gateways, L2 channels and memberships). Archived is a history record only. A subnet that is not bound to an active project is released. The other status changes do not touch the platform: blocking the organization's access, powering the project's virtual machines off, and deleting them while keeping a backup are not done by the portal today; the record and its review date are what it keeps.
11. Storage in the OpenNebula group¶
The group of a project (its ID: _<org>_<project>) carries an emergency copy of the project's current state as vector attributes; upper-case names, string values, empty fields omitted.
TENANT—STATUS,DATE_CREATED.QUOTA_<SITE>— per site:IMAGES_GBandFILES_GB(the sums of the site's pools),BACKUPS_GB(the site's one figure) and the site's connectivity, including the assigned subnets (INET_MAIN_SUBNET,INET_PUBLIC_SUBNETS,CORPNET_SUBNETS) and the L2 channels (L2_VNETS,L2_VLANS).QUOTA_<SITE>_<CODE>— per record:CPU,RAM_GB,GPU,DISK_GB(the pool's whole storage),DATE_CREATED.
TENANT = [ STATUS = "ok", DATE_CREATED = "2026-02-10" ]
QUOTA_DC1 = [
FILES_GB = "1", BACKUPS_GB = "500",
CONNECTION_TYPE = "inet",
IGW_TYPE = "vyos", INET_MBPS = "100", INET_MAIN_PREFIX = "31", INET_MAIN_SUBNET = "203.0.113.28"
]
QUOTA_DC1_CI1 = [ CPU = "10", RAM_GB = "20", DISK_GB = "100", DATE_CREATED = "2026-02-10" ]
Pending and archived projects exist only in the database. The portal does not compare this copy with its records: the consistency check of each platform (Site → Platforms) compares the quotas themselves.
12. Organizations, users, projects and the portal¶
- User: city — required, typed in as text; mobile number in international format (starting with "+"); passkeys (name, created, last used).
- Organization: type (cloud client or corporate), country, tax ID and its kind (the kind follows the country: IIN or BIN in Kazakhstan, TRN or trade licence number in the United Arab Emirates, a tax or registration number elsewhere; unique within the country), legal name — these four fixed once the organization is created — address and website (optional, editable), short name (3–15 characters; editable
only while the organization has no projects), sales rep, KAM, notes; registration (waiting / approved), city, the
resource manager's and the sales rep's approval of the registration (who, when); the registration check (score,
reasons, the address the request came from). Archiving an organization archives its users too.
- Registration (until the e-mail is confirmed): the registration check.
- SuperAdminRemoval: the super admin concerned, who asked, who confirmed, state (waiting / done / declined /
expired), when it was asked and closed; open 48 hours.
- TenantMove (a project moved to another organization): the project, the organizations (from, to), who asked,
the resource manager's and the sales rep's approval (who, when), state (waiting, moved, rejected, cancelled), reason.
- MailTemplate: an edited e-mail template (its name, text, who and when).
- Cluster: vCores per pCore — a whole number (left empty: 1 for a cluster whose hosts have GPUs, else 10); spare hosts N+X (left empty: 0 with GPUs, else 1); reserved CPU and RAM (%); for a GPU cluster the bundle size (vCores and GB RAM per card).
- Project (model Tenant): ID (short name, 4–15 characters; editable only until the project is provisioned),
description, type, stage, status, review date, created and days on the platform, project admin, support engineer,
resource manager, notes; the sales rep and KAM come from the organization. igw vCPU and RAM (GB); the internet
gateway VM as the platform reports it (name, image, vCPU, RAM, disks, when read). The internet gateway VM as the
platform reports it now includes its NICs (network, address); sizes in whole GB. The contract number and date and
the balance confirmations are gone.
- NetworkRequest: the site it belongs to (a project's connection per site).
- PortalSettings: provider name, logo (PNG), expiration notices switch; the timers (period, last run, state,
result) — PortalSettings.timers: the Registrations timer too.
A registration request that is not approved is deleted; only the portal's log keeps it.
13. Portal database¶
| Area | Main tables |
|---|---|
| People | users (staff and customers, roles, city, 2FA, passkeys), invitations, registrations, audit log, e-mail templates |
| Customers | organizations (type, country, tax ID, short name, sales rep, KAM, contacts), projects (model Tenant: ID, type, stage, approval, status, review date, order, OpenNebula group), memberships, join requests, project moves |
| Resources | quota records and connectivity (append-only versions), prefix assignments (address registry), network requests, L2 channels |
| Infrastructure | sites, IP subnets, server profiles, hosts, storage and volumes, networks with ranges and BGP peer groups, platforms with datastores, networks and clusters; core services and their nodes; secrets; templates; platform endpoints |
| Collection | observations and live statistics of OpenNebula |