Portal reference¶
How the Claugine portal works, how it is run, and what comes next. Addresses, host names, ports and accounts of the running installation are left out.
1. Overview¶
The portal is the single tool for day 0/1/2 operations of the OpenNebula cloud: customers register and order projects, staff approve them, and the portal keeps the authoritative record of organizations, projects, resources, connectivity and addresses. Tasks will also be callable through an API.
Each platform endpoint has a mode: read-only (the default: the portal reads OpenNebula, flags every difference with its own records, and writes only to its own test projects) or execution (the portal provisions projects on that platform). Every installation of Claugine starts from scratch.
Principles: the database is the desired state; OpenNebula is the actual state; nothing is ever deleted from the database (rows are versioned or blocked), except requests that never became projects and organizations without any history.
Conventions¶
What works the same on every page of the portal:
- A Copy button follows every field and every secret shown (a password once revealed, a key, a TOTP secret); a hidden secret has its own Copy button that copies it without showing it (each copy is logged).
- A small copy icon (⧉) follows addresses, names and other values shown as text, mostly in tables, to save space.
- Fields have one width across the portal; the copy control stays on the field's line.
- Sizes are whole numbers: RAM in whole GB everywhere; the vCores-per-pCore ratio of a cluster is a whole number.
- The portal says project everywhere: in the logs (actions such as
project_deployed, the sections and object types), in Tasks and in the names of the mail templates (project_ready_admin, …). A log search for "project" also finds the entries stored before the change. The wording changes nothing on OpenNebula: a project's group keeps its name (a group namedtenant-<name>stays so).
Lists, labels and states are described under Infrastructure.
2. Architecture and hosting¶
How the portal itself is built and run has pages of its own, under Running the portal:
- Portal architecture: the app VMs and their roles, MySQL with GTID replication, HAProxy on each site, the application (Apache in front of gunicorn, Django 5.2) and its timers, the HA agent, secrets, mail and certificates.
- Software delivery and updates: releases are bundles from my.claugine.com; each node downloads its edition's bundle and checks it; the standby first; automatic updates, the Updates tab, or over SSH.
- Backup and restore: daily backups to a local folder and an S3 bucket, the archive password, the platforms' front-end backups, a full restore and its checklist.
- Installation: the installation script.
In short: everything is in the database (users with their photos, organizations, projects, the infrastructure,
templates, tasks, logs and every secret, encrypted with FIELD_ENCRYPTION_KEY); only portal.env (the database
connection and that key) is kept outside it. Keep the archive password of the backups, or a copy of portal.env,
apart from the backups: without the key the stored secrets cannot be read.
The portal on two sites¶
Two app VMs at most, on one or two sites: one active (serves everyone, runs the timers), one standby (serves nothing; its database follows the active one by MySQL GTID replication; same version); maintenance serves nothing. HAProxy on each site asks each node for its role and sends the portal's name to the active node. The HA agent on each node runs the operations (switch-over, updates of the portal and of the OS, one node at a time) from Settings → Portal. Nothing switches by itself, except as a step of an update. Details: Portal architecture.
3. Users and roles¶
The portal has two kinds of users. Platform users work for the provider; organization users work for a client organization.
| Role | Kind | What it does |
|---|---|---|
| Super admin | platform | Every right in the portal: users and roles, settings, every approval and action. |
| Resource manager | platform | Approves new projects (names them, places them on a site and cluster, confirms the resources), resource changes, network changes and deletions; approves new organizations; edits organizations and projects; moves a project to another organization. Set per project (else the organization's, else any). |
| Support engineer | platform | Approves corporate projects when that step is ticked; sets the VLAN of L2 channels, changes the size of internet gateways, checks and makes again the provider's gateways, keeps the sites' IP subnets and address assignments, runs the platform work. Set per project (else the organization's, else any). |
| Sales manager (sales rep) | platform | Set on the organization: its manager, with full rights over it and all its projects (edit, order projects, add, change and remove their resources, delete projects, invite / lock / unlock its users, give the project admin role); adds organizations, becoming their sales rep; approves its new projects and its registration when sales approval is on, and a project moved to its organization; may move a POC's review date (up to 3 months after creation). Does not change the organization's sales rep, key account manager or resource manager, nor archive users. Sees the resource manager's dashboard for its own organizations. |
| Key account manager | platform | Set on a cloud organization: its account contact; the organization is read only to them except inviting, locking and unlocking its users and adding, changing and removing the resources of its projects; does not order or delete projects; gets its weekly report and the resource manager's dashboard for its own organizations. Optional. |
| Organization admin | organization | Everything of the organization: orders projects and asks for their changes and deletion, invites, locks, unlocks and archives its users, gives the project admin role; gets the access details of its new projects. |
| Project admin | organization | The projects given to them: asks for changes of their resources and networks; gets the project's access details when it is provisioned. Does not delete the project. |
| Read only | organization | Sees every project of the organization, changes nothing. |
A person may hold several roles; holding several, they approve once for all of them. Every role always has at least one active platform user (one who accepted the invitation and can sign in); the portal refuses a change of roles, a lock or an archive that would leave a role without one. A super admin has every right whatever roles are ticked for them; the roles ticked decide which approvals and notices come to them. Giving or taking the super admin role needs other super admins to agree within 48 hours (1 with one super admin, 2 with two or three, 3 with four or more, the one asking counting as one); a super admin can also act as any user who is not archived, read only. An organization's sales rep, key account manager and resource manager have full rights over its resources (the same actions as a super admin on its projects) and, from the organization's page, invite, lock and unlock its users; the sales rep and the resource manager also change their roles there. The rights of each role and these rules: Roles and organizations.
Administration → Users (super admin) has two tabs. Platform: email, name, state and a column per role (super admin, resource manager, support engineer, sales manager, key account manager); the requests to give or take the super admin role are shown on top, with who has confirmed and Confirm / Decline buttons. Organizations: email, name, state, organization, projects (one per line; * marks project admin) and the roles org admin (everything in the organization), project admin (only the projects given) and read only (sees every project of the organization, changes nothing); organization admins have the same list in their portal (Organization → Users), with the users asking to join on top. States: active (the invitation was accepted, sign-in works), blocked (accepted, sign-in locked), invited (sent, not accepted), inactive (not invited), archived. The bar has the tabs, a Filter box and an archived button. Roles are checkboxes in each row (ticking project admin asks for the projects; ticking or unticking SA opens a super admin request). Actions in a row's details: Send invitation / Resend invitation / Invite back (platform users get their roles with the invitation — no role before it), Lock / Unlock, Archive (no sign-in, no roles, kept for the record; users are never deleted), and Delete invitation for an invitation not yet accepted. The details show e-mail, mobile, organization, state, last sign-in, roles and projects.
Every user has a city, typed in as text (required in every form that creates or edits a user), and a mobile number written in international format, starting with "+" (the field offers it first; the portal checks the format). A super admin completes an organization user who has not accepted an invitation yet — e-mail, full name, mobile, city — and sets their roles (organization admin, projects) from Users → Organizations.
One email is either staff or a customer, never both. Each project is also associated with one OpenNebula user — the admin of its group — read at every collection.
4. Organizations¶
- The legal entity, with its country and its tax ID, chosen together when the organization is created; the
country, the kind of tax ID, the tax ID and the legal name are never edited afterwards. The tax ID follows the
country: in Kazakhstan the IIN or BIN (12 digits with a check digit; a number given without its kind is
checked against the kind it looks like, so one that is neither a valid IIN nor a valid BIN is refused at once), in
the United Arab Emirates the TRN (15 digits) or the trade licence (registration) number from the company's
licensing authority — an emirate's economic department or a free zone: 3–20 letters, digits and dashes, as written
(for example
CN-1234567) — elsewhere the country's tax or registration number (3 to 32 letters and digits). A tax ID is required for every organization, and it is unique within its country. The countries offered are set in Settings → Customer → Countries. An organization has no object in OpenNebula. - Address and website: optional, changed at any time on the Edit page; asked for on registration and carried
to the organization. A website given without a scheme is taken as
https. - Type:
cloud(external client) orcorp(internal client of the group). An organization is corporate when its admin registers with an email domain from the corporate list (Settings → Customer → Customers), or when staff make or invite it as corporate; the resource manager can correct the type when approving a registration. Once the organization exists its type, country, tax ID and legal name are fixed. Only corporate organizations may order corporate projects and the corporate network. - Invitations: the super admin, resource managers, support engineers and sales reps can invite a person to register a new organization (the Invite an organization page), choosing its type and sales rep (a sales rep invites for themselves). The person fills in the registration form with the email locked; no confirmation email and no approval are needed.
- Short name: 3–15 lowercase latin letters and digits, starting with a letter; suggested from the legal name (up to 8 characters, legal forms such as TOO or IP left out); editable by staff only while the organization has no projects (rejected or cancelled requests do not count) and nothing on the platforms uses it.
- Platform users of the organization: the sales rep (required for cloud organizations; optional for corporate ones, with an e-mail of a corporate domain from Settings → Customer → Customers) and the key account manager (optional; cloud organizations only), set on the organization's Edit page and applying to all its projects. The resource manager and the support engineer belong to the project (its Edit page); the resource manager approving a registration also assigns the organization's, used for its projects that have none.
- A corporate organization may have no users of its own: platform users of a corporate e-mail domain can act as its users — a corporate project's admin, and the organization admin (chosen on its Edit page), can be a platform user.
- The Edit page of an organization: its short name, address, website, sales rep, key account manager, organization admin (corporate organizations) and notes; the buttons (Save, Back) are on top.
- Organization users are invited by resource managers, sales reps and super admins on the organization's page (+ Invite user, with the roles) or added there (+ Add user: name, e-mail, phone and city, a record without sign-in whose row then has Invite; only while Settings → Customer → Customers → New users is on; switching it off removes adding, not inviting), invited by the organization admin, or register themselves. Every project of an organization with users has a project admin; when one is missing, an organization user is made its project admin (the organization admin first).
- The Organizations list (Cloud → Organizations): a Filter box, + new and archived on the title row; columns country (its two-letter code, the name as a tooltip), tax ID, organization ("registration waiting" for one not approved yet), type, users, projects, approvals waiting, manager (the sales rep; the resource manager of a corporate organization) and KAM; type, manager and KAM filter from their headers. Selecting a row shows its details — the name, the country, the tax ID and its kind, as in "ACME LLP AE 100000000000003 (TRN)", with the website and the address on the next line — and its type, with Edit, Archive (only when no project is left that is not archived), Unarchive and + Project; created and days on the platform; sales rep and KAM; then its projects (the name links to the project, with a copy icon), its users, its resource change history and its join requests.
- Support engineers and super admins can also be users of corporate projects (project admin or project user), added on the project page.
- Registering with a tax ID that already exists in that country never creates a second organization: it becomes a join request to the organization admins; staff are informed.
- Organizations are archived, never deleted (only a refused registration is removed); archiving an organization archives its users too (no sign-in, no roles, no open invitation), and unarchiving it does not bring them back.
- Organization area (left menu for customers): Dashboard — projects approved, deployed and waiting, resources of the quotas in force (vCPU, RAM, GPU, storage), VMs running/total from the last collection, users, join requests, open invitations and the latest events; Projects — the projects and ordering; Users (org admins) — users and their roles, join requests, invitations; Keys — the ssh public keys of the projects the person manages (add several at once, pasted or as .pub files, to several projects; assign them to the igw) and the organization's default ssh key; Tasks and Logs (org admins) — the tasks, and every operation on the organization, its projects and its users, filterable by project. Project users see the dashboard and projects they belong to.
5. Projects¶
Cloud → Projects lists the approved projects: on the title row a Filter box, + new (resource managers and the super admin) and archived; columns organization, project (with a copy icon), type, stage, status and review date (amber when close — a week for a POC, 3 months for a client — red when it has come; customers see the created date instead); type, stage and status filter from their headers. Selecting a row shows the project's page under the list. Requests that are not projects yet are under Administration → Approval requests. All activity is seen in Logs and Tasks — the same pages in the super admin's and the organization's portal.
The project page¶
Under the project's name (its ID, then its description): its type, stage, approval and status, and the date it was created. The first line of the details carries the name, the organization, the type, stage and status, and the actions: Edit, Request deletion and Resend credentials. The next line is the review timeline: the created date and the days since, a bar filled with the time already passed, and the days left to the review date (amber within 14 days, red with the days overdue). Then the organization (tax ID and name), the OpenNebula group and its id, the support engineer and the project admin; OpenNebula access (under a triangle) gives the admin and its id and the password — hidden, with a copy icon, Show (logged) and Reset.
Edit opens the project's own page (Save and Cancel on top): Move to another org…, the ID (read only), the type (changeable only for a project of a corporate organization), the stage, status and review date, the team (project admin and support engineer, with their e-mail, mobile and city) and notes.
Resource pools: one row per site and cluster — the site (a link to OpenNebula), cluster, vCPU, RAM, GPU bundles, storage with its split bar (VM disks, images, files) and backups, in use / quota as last read. Each quota change (an approved resource change, a resource pool ordered or removed, a deployment) sets the quota on the platform and then reads the project's group there again at once, so the pool shows the platform's new quota right after the change, not the figures of the last regular read. If that read fails, the figures wait for the next regular read of the Platforms timer (every 30 minutes by default); the quota change itself stands. Each resource pool shows its quota and, beside it, what is in use, with a bar that fills as the pool fills; the storage figure is broken into virtual machine disks, images and files: a pool's storage is one figure, images + files + VM disks, where images and files are set for each pool and VM disks are the rest, on the pool's cluster. The backup allowance belongs to the site, not to a single resource pool: it is one figure per site, shown and changed on the first resource pool of each site, and the change applies to every pool the project has there. How these figures become quotas on the platform: Storage quotas. The actions Storage (the split), Edit (ask for other resources) or Resources (staff), Remove (the request lists the pool's VMs and asks to confirm their deletion), and + Order resources (another cluster). Changes saved on the staff's Resources page ("Edit resources") apply to a deployed project's quotas and access rules on the platform at once; if the platform does not take them, nothing is saved. Pools are not removed there: a pool is removed through its removal request (Remove), which checks the VMs on it. A waiting change is shown as a row under the current resources. Below come the Internet, Corporate and L2 tables, the Gateways, the Users (+ add: support engineers and super admins of a corporate project) and the resource change history.
Internet. A project has one internet connection per site; its resource pools can be on several sites, so the Internet table can have several rows — one per site: the site, the speed, the internet gateway's type and the subnets (the main one first). The row of the project's own site has the actions: Speed (a new speed, for approval), Type (the internet gateway's type: changing it replaces the igw), Main (replace the main subnet, for approval — the dialog lists what happens after approval), + Subnet (an additional public subnet, for approval) and − Subnet (remove an additional one, for approval; shown when the row has one). Every request belongs to the row's site. The portal makes and changes a project's networks on the project's own site (the site it was ordered on); on the project's other sites the row says that changes are asked through support instead of offering the buttons.
Gateways. For staff the table shows each gateway, its VM name, management address and status (for the edge
gateway also compliant or not and the last check); the actions are Details, Check and Redeploy… (support
engineers and the super admin; the egw and cgw made again, and the gateways of an L2 channel) and, for a VyOS igw,
Edit. Redeploy… on an L2 gateway makes both sides of its channel again: the gateways are removed and made
with the standard names on the same VLAN and the same project network; the channel's traffic stops for a few
minutes. If a redeploy of the egw or cgw stops while the old gateway is being set aside (switched off, its address
moved, renamed), that is undone: the old gateway gets its address and name back and runs again. Each gateway's
details are compact: its status in the title; the VM (host name, platform, id, cluster, host); its size; its
NICs, one per line (network and address); for the edge gateway also its compliance with the standard, its login and
the last checks (✓ / ✗ per check, the failing ones with their detail).
The internet gateway's details also give the ssh command (ssh -p <port> -l <user> <address>) and its password:
hidden, with a copy icon, Show (logged), History and Change.
A VyOS internet gateway has an Edit button on its row: its configurations (read, store, apply with
commit-confirm, run commands), its networks and its size — vCPU and RAM; applying a new size restarts the
internet gateway's VM (the project's internet stops for about a minute), so the portal asks first.
The internet gateway is made with 1 vCPU and 1 GB RAM by default; what the platform reports for its VM (name, image, vCPU, RAM, disks and NICs) is read with every run of the Platforms timer.
flowchart TB
T["Project web · acme — cloud · poc · ok<br/>Edit · Request deletion · Resend credentials<br/>created · days · review timeline<br/>organization · OpenNebula group · support engineer · project admin"]
T --> P["Resource pools<br/>site · cluster · vCPU · RAM · bundles · storage · backups"]
T --> C["Internet"]
C --> R1["Site 1, the project's own site: speed · type · main subnet, additional subnets<br/>Speed · Type · Main · + Subnet · − Subnet"]
C --> R2["Site 2, another site: speed · type · main subnet<br/>Changes on this site: through support"]
T --> G["Gateways"]
G --> E["Edge gateway details<br/>status · VM · size · NICs one per line<br/>compliance · login · last checks ✓ / ✗"]
G --> I["Internet gateway details<br/>status · VM · size · NICs one per line<br/>ssh command · password: copy, Show, History, Change"]
I --> IE["Edit (VyOS): configurations, networks, size<br/>a new size restarts the VM"]flowchart TB
T["Project web · acme — cloud · poc · ok<br/>Edit · Request deletion · Resend credentials<br/>created · days · review timeline<br/>organization · OpenNebula group · support engineer · project admin"]
T --> P["Resource pools<br/>site · cluster · vCPU · RAM · bundles · storage · backups"]
T --> C["Internet"]
C --> R1["Site 1, the project's own site: speed · type · main subnet, additional subnets<br/>Speed · Type · Main · + Subnet · − Subnet"]
C --> R2["Site 2, another site: speed · type · main subnet<br/>Changes on this site: through support"]
T --> G["Gateways"]
G --> E["Edge gateway details<br/>status · VM · size · NICs one per line<br/>compliance · login · last checks ✓ / ✗"]
G --> I["Internet gateway details<br/>status · VM · size · NICs one per line<br/>ssh command · password: copy, Show, History, Change"]
I --> IE["Edit (VyOS): configurations, networks, size<br/>a new size restarts the VM"]
Moving to another organization. A resource manager asks to move a project from its Edit page (Move to another org…), choosing the organization from a list (tax ID first, sorted by name). The move waits for the resource manager and the new organization's sales rep (when it has one); the person asking approves in the roles they hold, and one person in both roles approves once. The request is shown on the project and among the approval requests; it can be rejected with a reason. When it is approved the project moves; its project admin stays when they can be one in the new organization, otherwise the new organization's admin (or another of its users) takes over. The steps are drawn in Approvals.
Project attributes¶
A project has its own resources, network, team and notes. The contract number and date and the balance confirmations are no longer kept.
| Attribute | Values |
|---|---|
| Type | cloud — external clients; corp — internal clients of the group (only corporate organizations may order it). The type of a project of a corporate organization can be changed by a resource manager or the super admin; a cloud organization's projects stay cloud. |
| Stage | poc or client, chosen with the order (POC by default) with its review date (by default a month later; a POC's at most 3 months later, except for resource managers and the super admin; the sales rep may extend a POC up to 3 months after it was created). Changed by the super admin, resource manager, support engineer or the organization's sales rep. |
| Connections | corp projects: internet, corporate network or L2 with either; cloud projects: internet, L2, or both. Internet and the corporate network are never in one project; L2 alone only in a cloud project. |
| Approval | pending → approved, or rejected / cancelled. |
| Status | ok, paused, suspended, saved, archived (see lifecycle). |
| Dates | created, days on the platform (from the first finished deployment), last modification, review date (the end of a POC or of the current status). |
| Origin | registered in the portal, or a test project. |
Analysis (Enterprise)¶
Analysis is a menu item below Projects in the Enterprise edition, for the super admin and the resource manager (all organizations), and the support engineer, the manager and the KAM (their own organizations). It has two tabs.
- Summary is the Dashboard's view (the tiles, clusters, storage, network pools) with three more tables: Platforms, right after the tiles; Backups (type, date, status, size); and Portal updates (date, version, standby node, active node). Backups and updates show the last 5 days; older ones of the period open with Show older. An update of a pair shows as one row per release, with the state of each node: first the node that was the standby, then the node that was active - also when the update stopped half way.
- Usage lists connection rows (type, speed, prefixes, IP addresses, the site's images, files and backups) and resource rows (pool, cluster, type, CPU, RAM, GPU, disk). A cell shows in use / allocated, the percentage on hover; Team, Connection and Resources switch row groups on and off without changing the filters chosen; columns filter and sort, and the project column stays in view when scrolling. The table can be e-mailed as PDF, CSV or XLSX. Names in the exports are always text: a name that begins with =, +, - or @ is never read as a formula by a spreadsheet (in CSV it gets a leading apostrophe). A site's backups quota is counted once per project and site (it is one figure for the site, whatever the number of pools there).
Platforms and the colours of capacity¶
The Platforms table has one row per platform: site, name, then VIP, Leader, Follower and Follower. Each cell shows the node's name with a coloured dot, and its address under it.
- The dot of the VIP is green when the VIP answered at the platform's last collection attempt, red when it did not answer (the VIP's address is not up on a leader), amber when the last attempt is more than an hour old, and grey when the platform was never collected. Errors of single requests during a collection are shown as a warning sign beside the VIP; they do not make the VIP red.
- A node's dot is green when it is the leader or a follower, amber while it is a candidate (an election is under way) and red when it is unreachable; the state is the dot's tooltip.
A cluster's CPU, RAM and GPU each have their own colour:
| Meter | Red | Amber | Green |
|---|---|---|---|
| Cluster, logical view | more is allocated than there is | above 85 % (CPU and RAM) | otherwise |
| Cluster, physical view | the allocation passes the usable capacity (the hosts less the spare hosts and the reserve) | — | otherwise |
| Datastore, the projects' quotas | over the usable size | above 85 % | otherwise |
| Datastore, OpenNebula's numbers | above 90 % | above 80 % | otherwise |
| IP subnet, delegated | above 90 % | above 75 % | otherwise |
The weekly reports colour their meters by the same rules (Reports).
6. Project ID (OpenNebula group name)¶
A project is one OpenNebula group, and the group name is the project ID shown everywhere in the portal. Projects are named _<org>_<project>, e.g. _acme_gpulab: the organization's short name and the project name without its hyphens. The limit in OpenNebula is 32 characters: organization short name 3–15 characters (lowercase latin letters and digits, starting with a letter), project name 4–15 (lowercase latin letters, digits and hyphens, starting with a letter), so _ + 15 + _ + 15 = 32 at most.
- The customer enters only the project name when ordering (the order offers
tenant1,tenant2, … free in the organization); the portal shows the full ID with the prefix. The project name is unique within its organization, with or without its hyphens. - The ID is fixed at the resource manager's approval (the manager may still correct the project name while approving, until its gateways are deployed); the portal refuses an ID used by another project or by an existing OpenNebula group. A rejected or cancelled request frees it.
- OpenNebula groups cannot be renamed, so a project ID never changes once its group exists, and the organization short name is fixed once the organization has a project.
- The project's other objects follow the ID: its OpenNebula admin
<org>-<project>-adm, its gateway VMs_<org>_<project>_<role>_<site>(egw, igw, cgw, …), its networks<group>-main_<site>.
7. Customer registration¶
On the organizations' portal anybody can register an organization. They give the organization's legal name, country and tax ID, optionally its address and website, and their own full name, e-mail, mobile number and city; they become its organization admin. The form offers the portal's home country first (Settings → Platform → Countries). The tax ID follows the country, as described in Organizations: in Kazakhstan the IIN or BIN, in the United Arab Emirates the TRN or the trade licence number, elsewhere the country's tax or registration number; it is required, and unique within its country. The city is typed in; the mobile number is in international format, starting with "+". They confirm the e-mail from the link the portal sends, and that sends the request. Nobody of the organization signs in until it is approved (the sign-in page says the registration waits).
- The resource manager approves first and assigns the organization's platform users: the resource manager and the support engineer (required; used for its projects that have none of their own), the sales rep (required for cloud organizations) and the key account manager (optional); the type (cloud/corporate) can be corrected here.
- When sales rep approval is on in Settings → Customer → Approvals, the organization's sales rep approves too (cloud organizations). A person holding both roles approves once.
- Then the organization admin gets an e-mail: the organization is registered, they can sign in and order resources.
A refused registration is removed: the person registering is told with the reason, the organization and its users are deleted (the log keeps the record). A tax ID already registered in that country becomes a request to join that organization, decided by its admins. Organizations made by staff, or registered from a staff invitation, need no approval. The steps and e-mails are drawn in Ordering and provisioning.
A registration request that is not approved is deleted; only the portal's log keeps it (e-mail, organization, tax ID, the checks' result):
- a request refused by the staff — at once;
- a request whose e-mail is not confirmed within 48 hours — by the daily run;
- a join request declined by the organization's admin — at once, with its applicant.
A request waiting for the staff's decision stays until it is decided.
The portal checks every registration — what is entered and how the form is sent. An automated submission gets the usual answer and nothing is made (the log keeps the attempt). Other signs of a fake or hostile request add to a score; a suspicious request is marked, and the staff see the score and the reasons on the organization waiting for approval and in the approval e-mail.
The deletions are made once a day (from 09:00) by the Registrations timer (Settings → Portal → Timers). There is no separate report on registrations: the super admin's weekly report counts the organizations, with the new ones of the period, in its Totals (Settings → Reports). The life of a registration request is drawn in Approvals.
8. Ordering a project¶
A project is ordered by an organization user on the organizations' portal, or by a platform user for an organization on the cloud portal (+ new in Cloud → Projects, or + Project on the organization). The order gives the stage and its review date, the site and resource pool, the resources (vCPU and RAM, or GPU bundles on a GPU pool; storage for VM disks, images and files together; backups), the connections (at least one), the project admin (a user of the organization) and a comment. Nothing is made on the platform by the order: it waits for approval, and provisioning starts by itself when the last approval is given.
| Field | Rule |
|---|---|
| Project name | 4–15 characters (lowercase latin letters, digits and hyphens, starting with a letter); tenant1, tenant2, … offered; the project ID is built from it (see naming). |
| Type | Cloud; corp is offered only to corporate organizations. |
| Stage, review date | POC by default, reviewed a month later (a POC at most 3 months later, except for resource managers and the super admin). |
| Site, pool | The site, then a resource pool (cluster) of that site. |
| Resources | vCPU, RAM GB, storage GB — defaults 4 / 16 / 100; on a GPU pool GPU bundles: one bundle is one GPU card together with a fixed share of its server's cores and memory, so a bundle always fits on the machine the card is in; the fields follow the cluster's type (see The resource fields); backups GB, 0 by default. A new pool has no images or files: all its storage is VM disks until the project admin splits it on the project page. |
| Connections | At least one of inet (internet), corpnet (corporate network — corporate projects only) and l2 (external L2-over-L3 channel). Internet and the corporate network never go together; L2 can be added to either, or ordered alone in a cloud project. Support assigns the L2 VLAN number. In the connectivity record the routed connection is stored (inet or corpnet), none when only L2. |
| Internet | Bandwidth 10–1000 Mbps in steps of 1 (default 100); the main prefix chosen from a list; + prefix adds an additional prefix, each chosen from the same list (− removes it). |
| Gateway | With the internet: advanced (the default), the platform's native router, or your own. With your own, nothing is created: the address is reserved and the project's virtual machine takes it. |
| Corporate network | Corporate projects only. The base subnet and the prefix length, both chosen from lists, as for the internet. |
| Prefix lengths | /31, then /29, /28, … down to the largest block the site's IP subnets allow (their max; see Network model); /31, /29, /28, /27, /26 while a site has no IP subnets defined. The customer chooses sizes; the portal assigns the addresses. |
A pending order can be cancelled by the organization admin or a resource manager; rejected and cancelled requests can be deleted. Cancelling or rejecting releases reserved prefixes and the group name.
9. Approvals¶
| Approval | For | When |
|---|---|---|
| Resource manager | every project | When its box is ticked in Settings → Customer → Approvals. Names the project, chooses the site and cluster (its platform and front end follow) and the internet address pool (set to auto when the ordered base subnet is not one the platform provides), confirms or changes the resources and the project admin, and picks the organization's sales rep when it has none. |
| Sales rep | cloud projects | When its box is ticked. |
| Support engineer | corporate projects | When its box is ticked. |
- In Settings → Customer → Approvals each approval step has a box. When it is ticked, every new order waits for that role: the project's or organization's own person in that role, else anyone with the role who can sign in, else a super admin. When it is cleared, nobody in that role is asked. Without the resource manager's step, an order is approved as it was ordered; if that cannot be done, the step comes back to a super admin with the reason on the project.
- The project's (or its organization's) resource manager approves when one is assigned and can sign in; otherwise any resource manager. A person in several roles approves once for all the steps waiting for them. A super admin may approve any step; they are asked only in the roles ticked for them.
- The resource manager's approval reserves the prefixes. When the last approval arrives the project becomes approved, its first quota and connectivity records are written, and its provisioning starts by itself (see Project deployment).
- Administration → Approval requests lists everything waiting — new organizations, new projects, resource changes, network changes, projects to provision, L2 channels waiting for a VLAN, deletions — with who approves now and "Waiting for me"; the menu shows how many wait for you. Approved and rejected requests of the last 30 days have their own tabs. A request can be rejected with a reason; the organization's admins are told. Moves of a project to another organization are listed there too.
- Changes and deletion: resource changes, network changes and deletions are approved by a resource manager (network changes also by a super admin); an L2 channel then waits for its VLAN, set by the support engineer. See Approvals for the full matrix of who can do what.
10. Network model and addresses¶
Summary; the full description is on the Network page.
- A shared front end is a shared platform; a secondary zone belongs to the front end of its primary (a secondary zone at a second site still makes one front end).
- Shared front ends are completely separate: no exchange (IX) network and no host may be used by two of them.
- IP subnets are defined per site, in the site's IP subnets tab (support engineers): the subnet and its prefix, the max allowed prefix — the largest block (smallest prefix length) a project may be given from it: at most the subnet's prefix + 2 and never bigger than a /24, by default that limit, and enforced when addresses are handed out: a request for a larger block is refused — and the routing: BGP (ix) (announced through the exchange), static gateway (the gateway, inside the subnet, is required) or organization (the organization routes it; the portal only delegates the block). The type is found from the address: a private range is corporate, a public one internet. An IP subnet itself cannot be changed once added: Edit changes only its max allowed prefix (not below a block already delegated from it) and, for a subnet not routed by BGP, its routing and static gateway. It cannot be deleted while an exchange network carries it or a prefix inside it is assigned.
- Public and corporate subnets are bound to an IX network: in the BGP settings of an exchange (ix) network of the site's Networks tab the IP subnets it carries are chosen from the site's subnets of its type (internet for cloud-ix, corporate for corpnet-ix). One IX network can carry several subnets of one type; an IP subnet is carried by one IX network only. Each IX network serves exactly one shared front end. A project's prefixes come only from the IX networks of the platform it is placed on. portal-ix belongs to the ext-managed platform and is never used for projects.
- The prefix lengths offered when ordering are /31, then /29, /28, … down to the largest max of the site's IP subnets of that type. Each public prefix also has a base subnet, one of the internet subnets of the IX networks of the chosen cluster's platform, or auto (cloud pools first); the chosen one is tried first, then the others. The resource manager can change the main prefix's base subnet (the internet address pool) at approval.
- Allocation takes the first free aligned block, largest blocks first, skipping everything in the address registry, in OpenNebula group metadata and in the address ranges of OpenNebula's virtual networks. For a /31 main prefix the second address goes to the internet gateway.
- The service networks of the inventory that lie inside a public pool of a shared platform (for example
cloud-services, the first /29 of each cloud-ix pool) run on the management clusters. They are recorded in the registry as platform services automatically after every inventory change and are never given to projects. - The address registry is in the same IP subnets tab: each IP subnet with its assignments (main, additional, corporate, reserved, platform services) and how much of it is used, the assignments outside every defined subnet, and overlapping assignments; support engineers record assignments and reservations, e.g. a /32 for the portal VM. A subnet assigned by hand is checked against every assignment at the moment it is saved, so it never overlaps one made at the same time, by hand or by the portal; if one appeared meanwhile, nothing is recorded and the page says so. Release frees only subnets assigned by hand. A subnet the portal assigned to a project is freed by that project's network change or its deletion, once its network is removed.
- Each routed connection has its own exchange network and project EGW leg: the internet EGW shapes bandwidth and passes only the project's public addresses (main public network /31 minimum, one address is the project's default gateway); the corporate EGW connects to the corporate network the same way. See the network reference design on the Network page.
- L2 (L2-over-L3 channel
s2s-NNNN, two provider-controlled legs bridging a client-site VLAN into a project VXLAN network): VLAN numbers are entered by support; projects bind a VNet (shown as "name (id)") to an L2 VLAN without approval, which reconfigures the VyOS VMs (with provisioning).
11. Resource records and metadata¶
- Storage: one number per pool for VM disks, images and files (a new pool has no images and no files: all of its storage is VM disks until the project admins split it), and backups apart (0 by default; one figure per site, shown and changed on the site's first resource pool, applied to every pool of the project there). Customers order storage and backups and ask for changes (Edit on the pool); the super admin and resource managers change the resources of an approved project on its page (Resources) — every change is a new version.
- Quota records per project, site and cluster (quota code): vCPU, RAM, GPU (fixed slices), storage (one number for VM disks, images and files; project admins split it on the project page without approval; until then all of it is VM disks), backups. Append-only versions with an effective date; a change is a new version, a removal is a version marked removed.
- Connectivity per project and site: connection type, gateway type, bandwidth, prefix lengths, L2 VLANs and VNets — also append-only.
- Full field list and the group format: Data model.
- The OpenNebula group template holds an emergency copy of the current state in the quota.md format:
TENANT(created, status),QUOTA_<SITE>(images, files, backups, connection type, bandwidth, subnets, L2) andQUOTA_<SITE>_<QC>(vCPU, RAM, GPU, storage), with the FireEdge views. Pending and archived projects exist only in the database.
12. Project lifecycle¶
| Status | Portal account | VMs | Network | Data |
|---|---|---|---|---|
| ok | active | running | working | kept |
| paused | blocked | running | blocked | kept |
| suspended | blocked | powered off | blocked | kept |
| saved | deleted | deleted | — | one full backup of every VM is kept |
| archived | deleted | deleted | — | nothing on the platform |
ok → paused (2 weeks) → suspended (1 month) → saved; back to ok from paused or suspended. The review date is the end of the current status. Statuses change by hand only (super admin, resource manager: Status on the project's Edit page); the one exception is saved → archived, done by the nightly round a month after saving. A saved project is archived on its review date, or earlier by hand (saved → archived). Either way, everything the project still has is removed first - on the platform (including the backups kept while it was saved) and in the portal (its subnets, quota records, L2 channels, memberships) - and the project is shown as "Being deleted" meanwhile; then it is archived and only its record stays. If something cannot be removed, the project stays saved, its page says why, and the daily round tries again. While a project is being archived, portal updates and OS maintenance wait for it, as for other running work. If an archival is cut off anyway (the node restarted), the project is saved again, its page says so, and the daily round archives it. Extending the review date of a saved project stops its archival if it is made before the archival starts; once a project is being archived, its review date can no longer be changed. 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.
Nothing is deleted: deleting a project or an organization archives it, and an organization can be unarchived. Every change of a project's resources is a new record; replaced versions are kept for the project history period (Settings → Portal → Keep history). The project page and the organization page end with the resource change history: date, action (create, update, delete), the project (on the organization page), the full set of values after the change, how long it was in effect (the set in effect now marked) and its status, newest first, with a Filter box. What happened to a project or an organization is in Logs.
Expiration notices go to the organization admins: before a production contract ends — 90, 30 and 14 days before, then every day of the last week — and while a POC runs out — every week, then every day of the last week. The notices follow the e-mail to organization users switch on Settings → Mail: with it off, a due notice is kept on the portal and logged (mail_held), never sent, so nothing is lost by leaving it off while the end dates are being set. The contract end date of a project is kept but no longer edited in the portal; the POC notices follow the review date. See Project lifecycle.
13. Infrastructure¶
- Lists work the same everywhere (as in FireEdge): a page shows 5, 10, 25 or 50 rows (the choice is remembered), a filter box searches the rows, the add button is at the top right, the actions of a row (edit, test, delete, …) are buttons on its right (in the organizations and projects lists they are in the details), and selecting a row shows its details under the list. Types are coloured labels with one colour per type everywhere (corp blue, cloud/client teal, poc violet, …); states are a dot and a word (green: active/approved, amber: planned/pending, red: retired/failed).
- The Infrastructure menu: Sites, Hardware profiles (tabs Servers and Boot storage), Templates, Secrets (users' public ssh keys are one of its types), then one item per site with tabs Platforms, Clusters, IP subnets (the site's IP subnets and the address registry, see Network model), Hosts, Storage, Networks, Core services. All infrastructure data are portal records edited only in the portal (list, details of the selected item, add and edit in dialogs); a record can be deleted while nothing refers to it.
- Site code: a number, required for a new site and fixed once set; it is part of the AS numbers of the site's networks (4 2xx xxx xxx is the private AS range).
- Management hosts and ports (site settings): addresses or host groups (e.g. 192.0.2.4, 192.0.2.0/28) allowed to manage the site's systems, and the ports for VyOS management ssh and for ssh and RDP to the management hosts. Every provider VyOS router (egw, cgw, L2 gateways, route reflectors — not the project's igw) has
eth0on the site'scloud-vyosnetwork in VRFMGMT; ssh listens only there, on the VyOS management port, and only the management hosts reach it. The other interfaces follow (egw: eth1 exchange, eth2 main network; cgw: eth1 exchange, eth2… corporate networks; L2: eth1 s2s, eth2 bridged). Ports: hosts, VMs and the rest use the standard ssh port 22, the portal VM too; only the jump hosts, reachable from outside, use non-standard ports for ssh and RDP; VyOS network functions use their own management port in VRF MGMT. The portal connects to every address of a site directly; the site's jump host is tried only when the direct connection fails. - Networks tab: type, interface, VLAN, MTU, subnet, gateway, ranges and BGP: local AS, the IP subnets it carries (exchange networks: chosen from the site's IP subnets of its type), then the peer groups rrs (addresses only), igw and egw (exchange networks: remote AS), each with hosts generated from a range and editable, and a password set from Secrets. Management interfaces of VyOS routers (VRF MGMT) get their addresses from the ranges automatically. 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. A network of type core-lm (one per site) carries 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 services tab: what the site itself runs, as a list. A service is deployed one of two ways: on hosts — on hosts that are already there (their OS), e.g. the OpenNebula front end of the management platform on each host of the management cluster, or NTP on those hosts; from scratch — VMs are made for it on a platform (usually the management one), all of the same size.
The list shows each service's name (a short name without spaces), its type (vm or host), how many hosts or VMs it has, the VM size (vCPU/RAM/disk), the image and the NIC 1 subnet, and its state. Its actions (Read, Check the definition, Edit, Delete) sit at the top of its details, as in the other lists. The details show the description, the platform hosting its VMs and the platform it serves (or "core" for a common service of the site), the definition — VM resources, the NICs with their subnets and gateways, routes, ufw rules — the secrets for ssh and for the service, the template, the checks, the VMs as defined (an address per NIC; "auto" when it is assigned at creation) and the deployed VMs as the platform runs them: name, host name, NIC 1 address, platform, cluster, host, status, with what each reported when read.
A service's form: its name and description; the platform hosting the VMs and the platform served (empty: core, a common service of the site); one row of VM resources — vCPU, RAM (GB), disk (GB), image, ssh port, NIC 1 subnet and gateway — and a row per further NIC (subnet, gateway; empty gateway: the subnet's); routes (NIC, subnet, next hop, metric); the VMs: name on the platform, host name and an address per NIC (empty: assigned when the VM is made); one secret for ssh and one secret for the service; ufw rules (port, protocol, from) for a service on a base OS — not for a VyOS image, whose firewall is in its template; the template, its differences and the checks. A VyOS service has the management network as NIC 1 and its own network as NIC 2.
Read (support engineers) reads the platforms' own VMs and logs in to every host and VM over ssh to note its host name, OS, CPU, memory, disk and addresses — read only, as a task; a VM whose size differs from its service's is flagged. Reading also checks every host and VM: the systemd services that must run, the ports that must listen, that its time is synchronized (NTP) and that all its BGP peers are established (IPv4 unicast or L2VPN EVPN), and that the addresses the platform gave a VM are on it. A service with a floating address (e.g. a front end in high availability) has a leader — the node that holds the address; exactly one node must hold it, and some services and ports are checked on the leader only (e.g. OneGate and OneFlow). The list shows how many checks failed; the detail shows each check per host or VM. The checks of a service are part of its form; the standard layout fills them.
Add the standard layout (super admin) makes the services of Claugine's standard layout that the site has, by the names of the management platform's VMs and by its management cluster; existing services are never changed by it.
Check the definition (support engineers; route reflectors) reads the platform and shows, for each VM, whether the service's definition describes it as it runs and as a redeploy would make it: NIC 1 on the management network, the address on the service network equal to the running VM's, no other NIC to carry over, the management address, the host name, the size, the image (customized through the context), the ssh port, the login user with an ssh key and its password kept from the running VM, the template that fits the network, the platforms. A redeploy starts only when the definition fits; the plan of each VM shows the same checks.
Redeploy… (super admin; route reflectors, from an image customized through the context). A core service's VM is rebuilt from its definition, never from the old VM: the name, addresses, password and configuration all come from the service definition, the site and the secrets. Redeploying it has four actions:
- Plan — shows exactly what would be done and changes nothing. Passwords are hidden.
- Redeploy — powers the old VM off, detaches it from the service's network so its address is free, and renames it
<name>-old; builds the new VM from the platform's image with the management network first and the service's network second, applying the configuration template as a one-time init script; then waits until it runs, answers over ssh, and has established its BGP peers. - Roll back — brings the old VM back, while
<name>-oldstill exists. - Remove the old VM — after the new one has proven itself.
If any part of building or checking the new VM fails, the portal rolls back by itself: the new VM is removed, the old one gets its network connection back and is resumed. A failed redeploy leaves the service running on the VM it was running on before. If the redeploy stops while the old VM is being set aside (switched off, its addresses moved, renamed), that is undone too: the old VM gets its addresses, name, autostart and scheduled actions back and runs again.
The site's own gateways — the EGW and the IGW of the services subnet — are built like the projects' gateways: from the egw and igw templates with the site's context (EGW: the exchange network and the services subnet as its client side; IGW: the services subnet outside, the core network inside, masqueraded) and the management part every provider router has (VRF MGMT). What a service has beyond that standard is its differences from the template: set and delete lines in its form, applied after the template (e.g. the IGW's further public addresses, the core services it publishes, routes to the internal networks). No secrets go there; they come from Secrets. Configuration (support engineers; on each VM of a route reflector, EGW or IGW service) shows, read only, the configuration a new VM would get, with passwords hidden.
-
Site settings: time zone, up to 4 NTP servers, up to 3 DNS servers for normal operation and up to 2 green-field DNS servers used when deploying from scratch while only the Claugine appliance is up.
-
Platforms tab: type — shared (projects are created on it), mgmt (the site's management platform), ext-managed (Claugine manages the platform itself — hosts, OpenNebula, networks — while its payload is managed by a third-party system) —, state (planned, deployed or retired, set by an operator; see States below), front-end mode (single, local, primary, secondary → its primary), the front-end address as a link to FireEdge (
http://<vip>:2616/fireedge/sunstone), operations mode (Switch to full mode/Switch to read-only: execution or read-only; only a deployed platform can be switched to full mode; logged). A shared platform has one public name, its Public FQDN (e.g.one-dc1.example.com), under which HAProxy publishes its front end (see OpenNebula front ends); FireEdge there is where projects open OpenNebula, shown on the project page (the site of each resource pool is a link to it).Add platform and Edit have Save and Cancel at the top and one row per option: Platform name, Type, State (Edit only), Mode with its primary (for a secondary), Runs on (VMs or the host OS), Zone name with OneGate through hosts, Secrets (super admins and the platform's), Public FQDN. OneGate through hosts ticked: VMs reach OneGate through their own host, OpenNebula's transparent proxy, at
http://169.254.16.9:5030; cleared: VMs reach it at the front end's VIP,http://<VIP>:5030, internal. The portal records the choice; the front end is set up by its operator to match. Below them, the front end's table, name and IP: the VIP (the front end's host name and VIP), then nodes 1–3. A platform always has all of them: the VIP and its host name, and each node's name and IP, every one required; each node has its own IP, other than the VIP. On a deployed platform only the state, the Public FQDN and the nodes' IPs can be changed. Each node's IP is kept with the platform; when one is empty, the portal takes it from the last refresh (OpenNebula's list of the zone's servers), matching the node's name; an IP set in the Edit dialog is never replaced by it.The details' header shows the platform, its type, its mode (full or read-only) and its consistency with the portal (consistent, inconsistent, not checked) with the time of the check, and the actions: Edit, Copy password (copies the oneadmin password to the clipboard without showing it; recorded in the audit log), Make consistent (when a repair is planned and the platform is in full mode), the mode switch and the + Datastore, + Network, + Image buttons. Below: the Public FQDN; the Endpoints — FireEdge, the XML-RPC API, OneFlow (
https://one-dc1.example.com/...) and OneGate for VMs: "from VMs, through their host" (http://169.254.16.9:5030) or the VIP's address, as chosen in Edit; the Clusters — the site's edge cluster and the platform's gateway cluster, set with Gateway cluster… at the top of the platform's details (a dialog); it cannot be changed once projects are deployed on the platform. The front-end table: the VIP and each node, with its IP, its state (VIP, leader, follower, candidate, unreachable, with a coloured dot: green working, orange in between, red not answering, grey not read yet), the last refresh and errors. Differences from the portal (with their number) are folded under an arrow: quotas, what archived projects left, groups without a project, public and corporate addresses, and the planned repair.Service images (purpose, type, name, id, status): three rows are standard —
vyosimage VyOS Router 2026.03 (egw, igw, cgw; the shipped default; only the name can be changed),one-routertemplate Service Virtual Router (the native igw; only the name) and its image (named as the template); other rows are free. Id and status are read from the platform by the health check. Each image has a customization — how a template is applied to a VM made from it: context — through a context variable at first boot (VyOS, like the projects' network functions); ssh — the template is turned into a script run over ssh (a base OS). The VyOS image cannot be deleted (network functions and VyOS core services are made from it; its name and customization can be changed). The native router's template and image can be deleted, together; the platform's projects can then only get a VyOS igw, and the health check stops looking for them. An image a core service uses cannot be deleted. L2 channels follow from the clusters, nothing is chosen: a shared platform with a dedicated edge cluster (its hosts carry the egw and the L2 bridges; one per platform, only on shared platforms) bridges a VLAN with one VyOS VM there, as the edge cluster is on the same EVPN fabric as the projects. Without one, the src gateway lives on the site's management platform and the dst gateway on the project's shared cluster, with a VXLAN tunnel between them. Read projects: when on, the Platforms timer (every 30 minutes by default) reads the platform's project groups with their users, quotas and templates, the VMs, hosts, address ranges and datastore usage into the portal (Dashboard, project usage, the consistency check below); it only reads. Off: the platform is not read (its health is still checked). On a deployed platform the front-end mode and its primary are fixed. The platform's name can be changed in Edit. Its key (site.name) changes everywhere in the portal — projects, deployments, collected data, a secondary's primary; the front-end (zone) name on the platform stays as it is. -
Consistency with the portal: at every refresh the portal checks three things on each platform, and the platform's details show the result under Consistency with the portal. Quotas: every active project has on the platform exactly the quota the portal records — vCPU, RAM, GPU and VM disks for each resource pool, the images, files and backups quotas on the datastores. Unbound objects: nothing on the platform without a project — a group belongs to an active project or to the platform itself; what an archived project left behind is listed. A group the portal does not know is shown but never touched. Addresses: every address range on the platform's public and corporate networks is in the address registry and belongs to the project that uses it; ranges registered to another or an archived project, overlapping ranges and registered ranges no one uses are listed. Platform-owned ranges are not checked. When something is not consistent, the details show each difference and the planned repair; a super admin can run all of it with Make consistent: the quotas are set from the portal (the portal is the reference), and what archived projects left is deleted the way deleting a project does it.
- Platform health: every platform is checked by the Platforms timer (Settings → Portal → Timers, default 30 minutes): the OpenNebula endpoint answers (its version is recorded) and, on shared platforms, the service images exist and are ready or used. The Platforms tab shows the result and the problems; Test checks at once.
- Hosts tab: IPMI address as a link (Redfish, always https; the IPMI https port of each host, default 443) and the IPMI type. IPMI settings reads every IPMI of the site through Redfish and compares it with the site: the same IPMI user (a credential), NTP and DNS servers of the site and its time zone (the IPMI user for every host of the site is chosen with IPMI user… at the top of Desired settings, a dialog); Apply to checked writes NTP, DNS and the time zone to the chosen IPMIs (some IPMI network changes need an IPMI reset; the IPMI messages are shown). The IPMI type found is written to the host file. An iLO holds two NTP servers: IPMI settings shows, applies and checks the first two of the site's NTP servers. The site's iLO syslog setting is compared and applied as one syslog subscription per iLO: an iLO with the old single-destination setting on, or with more than one subscription, shows as different from the site. Deploy, Reinstall and Retire on a host: see Deploying and retiring hosts.
- Disk sizes: the boot disks of server profiles are in GB (109 bytes), as hardware vendors give them — the only place with GB; volumes and everything else are in GiB (230 bytes).
- Reading the hardware of a host asks its IPMI first (Redfish): chassis, CPU, sockets and cores, RAM, GPUs, the boot volume (the boot controller) and the network cards with their PCI slots, models, ports and speeds. The names of the components come from there. The Linux names of the ports are predicted from the slots as systemd gives them:
ens<slot>f<function>np<port>(slot 1 →ens1f0np0,ens1f1np1; OCP slot 22 →ens22f0np0; a card in PCI domain 1 →enP1s2f0np0; a one-port card has nof0:ens3np0). The OS is then read over ssh (port 22) for what only it knows: the boot device and its volumes, and the aliases the ports carry (nic-data1 = ens1f0np0). Hosts in no cluster may not answer over ssh: that is no error. -
Hardware profiles has two tabs, Servers and Boot storage. Servers lists the server profiles: Check reads every host of a profile and compares (cards are paired by PCI slot; the aliases of the hosts with the profile's); it also takes the component names from the IPMIs (the name most hosts give). Port aliases: a server has 2 or 4 ports — two fabrics:
bond0=nic-data1+nic-data2(data: VMs, host to host, MTU 9100) andbond1=nic-storage1+nic-storage2(storage, MTU 9000); one fabric:bond0=nic-data1+nic-data2carries both (e.g. gi2 until its second card comes). The profile assigns each alias a port of its cards. Verify: a profile can be edited until it is verified (its aliases assigned, its hosts checked and the same); then only its volumes can change, and the aliases while no active host of a cluster uses it. FC HBAs are not used yet.Boot storage lists 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.
A server profile names its boot storage (it must fit the profile's boot device; a profile made from a host gets the first that fits) and, once its hardware is verified, the provider's part of the profile: the BIOS and iLO firmware versions a host must have to be deployed, the BIOS settings it gets, and the iLO services left on (ssh, IPMI over LAN). BIOS settings: the profile lists its own, one "Attribute = Value" per line. Two more rows are shown read-only and are always set by every deployment: "PXE boot: disabled on all NICs" and the time zone (the site's); they cannot be entered in a profile. The default profile settings are: WorkloadProfile Virtualization-MaxPerformance, AutoPowerOn RestoreLastState, PostF1Prompt Disabled, PowerOnDelay NoDelay, ExtendedMemTest Disabled, MemFastTraining Enabled, EmbeddedUefiShell Disabled, EmbeddedIpxe Disabled.
-
States: a platform's state is set by an operator on its form: planned, deployed or retired (see Inventory, States); changing a deployed platform's clusters, datastores or networks in Infrastructure leaves it deployed. Only a planned platform can be deleted; an installed one is retired instead. Items of a platform that is not installed are planned; installed items are active; removing an active host or cluster marks it retired until it is taken down, removing a planned one deletes it. Only active hosts give capacity and only active shared clusters of an active shared platform can be ordered.
- Clusters tab: a cluster row shows its platform, name, type, state, and what it holds: hosts, networks, datastores and the projects that use it. Selecting a row shows its details. A GPU cluster also shows its GPU model. Its capacity settings and the two views of its capacity are below (Cluster capacity). A planned cluster can be deleted. One in service is retired instead.
- Storage tab: storage is recorded in two levels. A storage row is the system itself: its name, type, the host it is reached on, and the volumes on it. A volume row is one exported volume: its name, device, total size in GiB, its export and targets, the clients that may mount it, the fstab parameters it is mounted with, and what it is used by. Both are added, edited and deleted from this tab. Deleting a storage system deletes its volumes with it.
- Entering the inventory: a portal is installed from scratch and starts empty; every item of the inventory, and every organization, project and person, is entered here, through these pages. Nothing is imported from another system or from files. The order and what each item is for: Inventory.
-
Cluster capacity: every cluster has its own capacity settings: vCores per pCore (how many vCPUs one physical core gives; hyperthreads are not counted), spare hosts (N+X: the largest X hosts are left out of what can be ordered; no host is a dedicated spare) and the reserved CPU and RAM percentages. The cluster's details show them as "10 vCores/pCore · N+1 · reserved CPU 5%, RAM 5%", for a GPU cluster too. In Add cluster, vCores per pCore and spare hosts start empty: left empty, a cluster whose hosts have GPUs gets 1 and 0, any other 10 and 1; a figure typed in is kept. Edit shows the cluster's own values, the same fields for a GPU cluster as for any other.
Two views of capacity. Logical (what projects can order): vCPU = (the hosts' cores − the spare hosts' cores − the reserve) × vCores per pCore; RAM = the hosts' RAM − the spare hosts' RAM − the reserve; GPUs = the cards outside the spare hosts. Physical (the hardware): the hosts' cores, RAM and cards, with the allocation as cores (allocated vCPU ÷ vCores per pCore); its bars show used, spare, reserved and free.
A GPU cluster's reserve percentages count in the physical view only; what projects can order in it is counted in bundles. A bundle is one GPU card with a fixed share of its host's cores and RAM, set with the cluster's Bundle size button (vCores and GB RAM per card, and an optional Storage per GPU bundle, GiB; the dialog says what the hardware suggests) and shown as its own row, "Bundle size", under GPU in the details (e.g. "7 vCores + 112 GB RAM per card + 500 GiB storage"). When the storage per bundle is set, a resource pool on that cluster has storage = bundles × that size: the storage field follows the number of bundles and cannot be changed in the order or change forms; when it is empty, the pool's storage is any size. When a bundle has more vCores than a card's share of the physical cores, vCores per pCore says by how much: for example, 4 cards on a host of 64 cores with bundles of 32 vCores need 2.
Deploying and retiring hosts¶
On a site's Hosts tab, Deploy (a planned host) and Reinstall (a host already in use) install the host from scratch: its iLO and BIOS set as its profile says, its local storage made again, Ubuntu installed from the site's portal node, then OpenNebula, its networks and datastores, and makes the new host the same as the hosts already in service. The host joins the chosen cluster disabled; enable it yourself after checking it. The dialog asks for the cluster, Reset to factory (a real BIOS factory reset first; the iLO keeps its address and accounts), the confirmation that a powered-on host may be powered off, and the host's name typed in full: its data is erased (the boot device partitioned again from the profile's boot storage, the volumes of the other controllers deleted with their data, the new datastore volume wiped). A step-by-step walk-through: Deploying a KVM host.
Before anything is changed the run checks: the platform is on its baseline (see Platforms), the internal domain is set (Settings → Customer → Customers), the host's profile is verified and its boot storage fits, the installer is on the site's portal node, and the host is in no other cluster in OpenNebula and has no VMs there (an existing registration is disabled at once, and checked again right before the power-off).
The run's steps and its log are shown on the host; a step that fails stops the run with its reason, and Resume from the failed step continues it (a run interrupted by a restart of the portal is resumable too); Abandon leaves the host as it is and deletes its answers ISO. Only one run per host at a time. A run resumed after an upgrade of the portal redoes the OS baseline, the packages, oneadmin's keys and the reboot when an older version did them, and sets LLDP on the iLO if it was not set yet.
Some steps in more detail:
- iLO settings: the iLO is named "m" plus the host's name without its leading "h". It gets the site's NTP servers - an iLO holds two, so the first two of the site's list - its DNS servers, its time zone and, when one is set, the remote syslog server: one syslog subscription (UDP, the site's syslog server), with the old single-destination setting off. LLDP is turned on, set and read back. If the iLO refuses the remote syslog setting, the run notes a warning in its log and goes on; every other iLO setting that is not applied stops the run.
- BIOS settings: after the reset to factory settings, the deployment sets only the server profile's BIOS settings, turns network boot (PXE) off on every NIC the BIOS has, and sets the site's time zone. Every other BIOS setting stays as the factory reset and the workload profile make it - with Virtualization-MaxPerformance, the workload profile itself decides settings such as PCIe ASPM and the NUMA options. After the installation the readback checks exactly the settings the deployment set. A deployment started by an earlier version and resumed after the update has its BIOS settings set again for the new version before the readback; if anything had to change, the run stops and asks to restart the host and resume.
- Local storage: on every controller but the boot controller (the one iLO names as the boot controller, holding the boot-size volume - nothing is deleted when that is not certain), the volumes are deleted and one volume is made over all its drives (RAID5 for three drives or more, RAID1 for two); on the host it becomes the volume group vg_one, mounted on OpenNebula's datastores folder. Deleting the old volumes loses everything on them; afterwards only the new volume, identified by its own identifier, is wiped. The boot device is partitioned again from the profile's boot storage when Ubuntu is installed. HPE iLO 6 and 7 know a server's storage only after the server has started (POST): with the server off they list none. So the step starts the host once into the BIOS setup - no operating system starts - waits until iLO has read the controllers, does its work, and powers the host off again, whatever the outcome, before the installer starts. The BIOS settings staged before are applied by that start-up. If the host cannot be powered off, the step fails and the run stops there.
- Ubuntu installed: the step waits until the host answers over SSH with the operating system this run installed and cloud-init has finished its first start. The Ubuntu installer turns cloud-init off after that first start; a host where the installer did so, and where cloud-init finished, counts as installed. The answers file's cloud-init part is as small as it can be: the host's name, root and password login off, and the SA user with its key - no other user and no packages. The installer itself only adds the SSH server.
- OS baseline: oneadmin is made after the installation, by this step: the oneadmin group and user are created with their ids and home given in the command (group 9869, user 9869, home /var/lib/one) and checked. This step does not create the /var/lib/one directory - the OpenNebula packages do, after it. oneadmin's sessions use umask 0027 (checked over ssh as oneadmin), so what OpenNebula's drivers create gets no group write and no access for others: a file created 0666 is 640, a directory created 0777 is 750. No kernel dump (kdump-tools, kexec-tools and the crashkernel option removed); time from systemd-timesyncd with the site's NTP servers (chrony removed); snapd removed and fwupd masked.
- Packages: lldpd, ufw and AppArmor; oneadmin in the disk, kvm and libvirt groups. The Packages step installs nothing unless oneadmin is 9869:9869 with home /var/lib/one, so no OpenNebula package is ever installed before it.
- Storage: /var/lib/one/datastores and every datastore in it are 750 oneadmin:oneadmin.
- FRR: the host's FRR configuration is the site's host-frr template (see Templates), filled in with the host's name, its address on the EVPN network the host uses and that network's route reflectors, AS number and the hosts' BGP password (Infrastructure → Sites → the network's peer groups; the password itself is kept in Secrets). The step stops before writing anything when a value is missing - the password included - and waits until every route reflector has accepted the session.
- Firewall, AppArmor, LLDP, a step before the reboot: AppArmor on; lldpd on; ufw (IPv4) dropping incoming and routed traffic except from the site's management hosts, the core, live-migration and EVPN networks, and ssh and the node exporters from the front end. VM traffic on the host's bridges passes ufw (OpenNebula's security groups filter it) and outgoing connections are open, so when the platform reaches OneGate through hosts (OneGate through hosts ticked) VMs still reach it through their host (see OneGate): OpenNebula's transparent proxy answers on each bridge in its own network namespace and connects out to the front end. While ufw is turned on, a way back turns it off again after a few minutes unless the portal can still open a new ssh connection to the host - so a wrong rule cannot lock the host out.
- Once OpenNebula monitors the host: RESERVED_CPU is set from the cluster's vCPU:pCPU ratio and read back; the host's IPMI health is read at the end of the run.
Retire on an active host in a cluster (super admins) opens a page that reads OpenNebula. When the host has local storage, it lists in red what is deleted with no way to restore it: the datastores on that storage, the images in them, and every VM with a disk there (on any host) - these VMs cannot be moved off it. To start, tick the deletion (when there is anything to delete) and type the host's name. The run, shown step by step on the host:
- R0 checks - the host is in OpenNebula and in its cluster, another host of the cluster is enabled, every VM is running or powered off and each fits on one other enabled host, the iLO answers, the local storage is what was confirmed and no other platform's datastore uses it; nothing changes if a check fails.
- R1 the host is disabled in OpenNebula. R1a the confirmed VMs on the local storage are deleted (each read again just before; a VM no longer there stops the run).
- R2 the other VMs are moved one at a time to the enabled host with the most free memory - running ones live, powered-off ones cold; the run stops at the first VM that does not move.
- R3 no VM is left. R3a the images and R3b the datastores on the local storage are deleted.
- R4 the host is deleted from OpenNebula.
- R5 the server is shut down through its iLO (forced after 10 minutes), only while the host is not in OpenNebula again and no datastore uses its storage.
- R6 the host is marked retired, its storage retired and its local datastore records removed.
A failed run is continued with Resume (from the failed step) or given up with Abandon. Once a run has begun deleting, only its own Resume continues it; the host shows what it was confirmed to delete. A walk-through: Retiring a KVM host.
The last run in a host's details: the last deployment or retirement of the host folds under its heading (open while it runs or after it failed). Its log shows only what the steps do not - the work inside them, such as each VM moved.
Secrets¶
Infrastructure → Secrets keeps every secret of the infrastructure, by type, encrypted with FIELD_ENCRYPTION_KEY. Lists show whether a value is kept, never the value; the super admin can show the values of a secret (logged). Each row counts its contributors (what uses the secret: platforms, hosts, sites, BGP peer groups, settings, the portal itself); used by N · show contributors opens the list. In edit dialogs a kept value is shown as set (●●●●●●●●), never its content; focusing the field shows the real value (logged). The copy icon next to a kept value in the details copies it to the clipboard without showing it (logged). Contributors refer to a secret by its id, so a secret can be renamed at any time; every reference to a secret (or to another object) is chosen from a list of the existing ones. The portal's own key on project routers is made on first use.
- Users — name (unique), status, user name, password and/or ssh private key (one of them is required), purpose, the public key calculated from the private key. Platform super admins, IPMI users, jump hosts and the portal's own service user.
- BGP — one per exchange network of a site, named
<site>.<network>, type ix or evpn: the passwords of the route reflectors (rrs), of the internal peers (igw: project gateways, KVM hosts) and, for ix, of the external peers (egw). All required. - OpenNebula — one per front end (e.g.
cloud.one): the OS useroneadmin(password of at least 30 characters and an ssh private key), the database (defaultonedb, useronedbuser, password of at least 30 characters) and the OpenNebula super adminoneadmin(uid 0; password of at least 15 characters). - Keys — public ssh keys of users (type, key, comment). A key cannot be changed after adding.
Status: new (added, not used), used, replaced (its contributors were moved to another secret, but a system may still accept it; it cannot be given to new contributors), archived (not used, kept for the record; Show archived). Replace points every contributor to another secret of the same type and starts a check (Tasks): the new secret is accepted and the old one refused — OpenNebula over XML-RPC, super admins and jump hosts over ssh, IPMI over Redfish; BGP and the rest are confirmed by hand. When every check passes the old secret is archived; otherwise Check systems runs the check again. Archive takes only a secret nothing uses. The portal does not change secrets on the systems: change them there first, then replace them here.
Settings → Backups → Platforms, Connect to Front-Ends using, can use a secret too (when one secret is chosen there): such a secret cannot be archived while it is chosen, and replacing it moves the setting to the new secret, which must have an ssh private key.
Templates¶
Infrastructure → Templates holds the sources of the configurations of network functions, taken from the running services: egw (internet edge: BGP to the internet exchange route reflectors, martians filtered, bandwidth limit), igw (project gateway: EXT main network, INT private networks with SNAT, PUB extra public networks, ssh key login), cgw (corporate edge), L2 channel project and external sides (bridge + VXLAN over the s2s network), the route reflectors of the internet, corporate and VXLAN/EVPN exchanges, and FRR of the KVM hosts. Templates leave out what the first boot of the VyOS image already sets from the OpenNebula context (addresses, MTU and IPv6 options of the NICs, default route, host name, time zone, login user and key, ssh basics, system options, syslog); only descriptions are in double quotes and only passwords in single quotes. The deployment passes the rendered configuration as the one-time init script ONE_TIME_INIT_SCRIPT_BASE64, which vyos-postconfig-bootup.script runs once on the first boot (commit and save; a failed command is reported by its line number). Templates use Django template syntax; addresses, names, ASNs, peer ranges and users come from the inventory and the project, secrets from Secrets. Preview… at the top of a template's details (a dialog with the project, or the site and the router) renders the template for a project or for a site and route reflector, with secrets masked as «secret:…» and values decided at deployment as «…»; the super admin can show it with secrets (audited). The project page links the configurations of its egw, igw, cgw and L2 sides. Reset restores the built-in version.
The host-frr template is the FRR configuration a host deployment writes on a KVM host (see Deploying and retiring hosts). It may hold plain substitutions ({{ name }}, {{ ix.asn }}, …) and for / if blocks only - no filters and no other tags. A host deployment checks this before it writes anything; a template that does not keep to it stops the FRR step with the line numbers and the reason (never the template's text), and is corrected here. Every credential the template uses must have a value in Secrets, or the step stops naming it.
14. Project deployment¶
Projects are deployed on platforms switched to execution mode (site → Platforms → Switch to full mode); on a read-only platform the approval says so and the project page shows why it was not deployed. The order offers a project name (tenant1, tenant2, … free in the organization). The igw ssh port is random, from the range in Settings → Security → igw ssh ports (50000–59999 by default, always within 10000–59999), chosen with the order.
Passwords (portal users, projects' OpenNebula admins and routers, secrets of the infrastructure) follow the password policy in Settings: by default at least 10 characters with 2 small letters, 2 capital letters, a figure and a special symbol; special symbols only from the allowed list (default - _ . + = ^ ~). Any other special symbol can be allowed, except those that break passwords where they are used: the space, quotes (' " `), ?, /, \ and |. A password may not contain a part (4 characters or more, 6 for figures) of the personal data of the person, the project or the organization: names, email, phone, tax ID (a switch of the policy). The last 5 passwords may not be used again (3–1000). The rules are shown under every new-password field and checked while typing. The project's admins can show the project's OpenNebula admin password (logged) and change it — OpenNebula is changed too. The passwords the portal makes itself (for example, for a new project's accounts) always meet the password policy: its length and the number of lowercase and uppercase letters, digits and special symbols it asks for. A policy that asks for special symbols but allows none cannot be met; the portal says so instead of making a password.
When the last required approval is given, the project is provisioned at once, with the defaults: the OpenNebula
admin <org>-<project>-adm with a generated password, the internet gateway (VyOS unless another type is chosen) with
its user and a generated password (no ssh key: the gateway accepts password login), its size (1 vCPU, 1 GB RAM by
default) and the project admin. There is no Provision button any more; Provision again appears only when the
provisioning could not start (the platform is read-only, for example) or failed — it goes on from what was made.
The deployment runs in the background as a task; its steps (pending, running, ok, failed) are in Tasks. A task cut off by a restart is marked failed ("interrupted") and no longer keeps updates and key changes waiting. The project admin gets the access details, with the platform user who provisioned the project in copy; every organization admin who is not the project admin gets the same e-mail of their own. Only when the project admin's e-mail went out are the project's platform users told. "Resend credentials" sends it to the project admin only. New users of a project get their access details by e-mail as soon as the project is provisioned, without a separate invitation. See Ordering and provisioning. Before creating anything the main prefix is checked against every address range in OpenNebula; collection records those ranges and the address allocation skips them.
Every operation on the project — order, approvals, prefixes, each deployment step, password views and changes — is in Logs (the super admin's, and the organization admins' Organization → Logs).
- Group
_<org>_<project>on the primary zone of the federation, with the template copy (TENANT,QUOTA_<SITE>,QUOTA_<SITE>_<QC>) and the FireEdge views. Every project's group has FIREEDGE=[DEFAULT_VIEW="user", GROUP_ADMIN_DEFAULT_VIEW="groupadmin", GROUP_ADMIN_VIEWS="groupadmin,user", VIEWS="user"]: its users open FireEdge in the user view, its group admin in the groupadmin view with the user view to switch to. The portal writes it when it creates a project's group and each time it applies the group again. - Admin
<org>-<project>-admin the group, admin of the group. - Quota as in
lib/claugine_quota: the project's cluster gets VMS and CPU twice the running limits, MEMORY twice RUNNING_MEMORY, SYSTEM_DISK_SIZE, PCI_DEV = GPUs; the other clusters -1; datastores: SYSTEM_DISK_SIZE is the VM part of the storage; the image datastore of the pool's cluster gets the images part (pools whose clusters share an image datastore share its quota, their sum), and the platform's other image datastores are closed to the project (0); the site's file datastore gets the files part of all the project's pools there (their sum); the backups figure of the site goes to the platform once (see Storage quotas). The quota is set on the group in every zone where the project has a pool (see Quotas in every zone). - ACLs as in
lib/claugine_aclwithout role groups (as on the live platforms): the group admin holds the super admin rights, the group the user rights, in the platform's zone; encoded to the hex form ofone.acl.addrule(checked against the rules of the live platforms). - Main public network
<group>-main_<site>: VXLAN/EVPN, MTU 1500; the lowest address is the egw and the gateway, the next one the igw; DNS of the platform's services network. - egw
_<org>_<project>_egw_<site>: provider's VM in group oneadmin, 2 vCPU / 4 GB, NIC 0 on the management network (VRF MGMT), NIC 1 a free address of the exchange network (cloud-ix), NIC 2 the egw address; user and key of the platform's super admin (from Secrets). - igw
_<org>_<project>_igw_<site>: the project's VM (owned by its admin), 1 vCPU / 1 GB RAM by default, NIC 0 the igw address with the egw as gateway; its user with a generated password.
Both routers come from the platform's network function image (default VyOS Router 2026.03, set per platform); the VyOS context sets addresses, user, key and time zone on the first boot; the rest of each router's configuration is built from the templates (see Templates) and run once at that first boot. Every step is idempotent: an object that exists by name is reused, so a failed run can be repeated. A network is reused only when it is the project's (and an older additional prefix's network only inside its prefix): one that only has the expected name is reported as left alone and the igw is not attached to it. The steps and what they made are in the task (Tasks).
Deleting a project: when the deletion is requested, every resource of the project on its platform is listed and attached to the request — VMs including the egw, igw, cgw and L2 gateways (every provider gateway — egw, cgw, L2 gateway — that holds an address on one of the project's networks is listed, whatever its name, and terminated in the VM step, before any of the project's networks is deleted), backup jobs, virtual routers, templates, images and backups, security groups, VM groups, networks with the main one, ACL rules of the group and its users, the users and the group (a resource is the project's when its group is the project's or its owner is a project user). After approval exactly the listed resources are deleted; before the users and the group go, and again at the end, the platform is checked: if anything of the project is left (for example a backup made after the request) the deletion stops and lists it, and a new run lists everything again. Org admins, the organization's sales rep, resource managers and the super admin request it on the project page (typing the project name to confirm; not a key account manager or a project admin); a resource manager of the project (or the super admin when the RM approval is off) approves it there or in Approval requests → Deletions. Resource managers and the super admin can also delete a project at once (Request deletion on the project page). The deletion removes everything of the project in OpenNebula, in this order: its L2 channels, on both ends (with the gateway and VLAN network on the site's management platform); its VMs together with the provider's egw/cgw/L2 gateways on its networks (terminate, then wait); backup jobs, virtual routers, templates, images, security groups, VM groups and the project's networks; a check of the platform; the ACL rules of the group and its users, users whose primary group is the project's (others only leave the group) and the group; a check again — then the portal releases its prefixes, closes its quota records, removes memberships and marks it deleted; the record and its log stay. On a read-only platform or for a project never deployed only the portal part runs. A failed run can be repeated: what is already gone is skipped. If a deletion fails half way (the platform did not answer, for example), the project stays deleting. A resource manager or a super admin then has Retry deletion on the top row of the Delete card: after typing the project's name, what is left of it on the platform is listed again and deleted; what is already gone is skipped. Every object gets its own status on the request (deleted, gone, left the group, failed) and a line in the project's log. What would be deleted shows the plan without writing. When the platform cannot be read during a project's deletion, the deletion stops at that step (it can be run again); it never goes on to the permissions, users and group without knowing what is left. When a project has resource pools in more than one zone of the federation, its deletion lists, deletes and checks its resources in every one of them; the project's users, group and access rules go only when nothing of it is left in any zone. A zone that cannot be reached stops the deletion before anything is deleted; it can be run again. A project that was deployed on a platform the portal has no control or no credentials for is not deleted: the deletion stops before anything is touched, with the reason in its plan, and nothing of the project is released until the platform is configured again. A platform in read-only (adoption) mode, and a project that was never deployed, are deleted from the portal only, as before.
Configuration check (the VyOS timer, every 480 minutes by default; Check on the Gateways table checks that each gateway answers ssh and its BGP peers are up): the portal compares every VyOS router of the provider with the configuration built from its template — the sites' route reflectors and egw/igw, and the projects' egw, cgw and L2 gateways. A project's igw is not checked: after its deployment it is the project's own. The portal reads show configuration commands and lists the lines missing on the router and the lines the router has under the paths the template manages that the template does not give, including a line the template deletes (such as a default route in the main table). The tunnel keys of an L2 gateway are made with the gateway and kept nowhere, so they are not compared. The timer's result names the routers that differ from their template and those that cannot be reached; each router's latest result is kept with it. Quotes are ignored. Every credential in a stored line (a password, a key, a secret) is hidden whole, whether or not the portal knows its value.
SSH keys of a project (Organization → Keys): project admins add several public keys at once and assign them to the igw; its configuration lists every assigned key. Without a key the igw accepts password login. For a running igw the page shows the commands that make it match the assigned keys.
Networks of a project (project page → Internet, Corporate and L2; one internet connection per site, every request belongs to its site): project admins ask to add a public subnet (/24–/32, with its base subnet or auto), a corporate network (corporate organizations only) or an external L2 channel (an internal L2 channel, between two of the project's networks on two sites, is added without approval); to delete an additional public network, a corporate network (not the last one) or an L2 channel; or to replace the main public network. A resource manager (or a super admin) approves in Approval requests or on the project page (an Internet or corporate-network request with Approve…, which asks for the delegated range when there is one to choose); an L2 channel then waits for its VLAN, set by the support engineer; the change is then made at once on a deployed project whose platform is in execution mode, step by step, with each step on the request (on the project page, with Retry after a failure; the steps are in Tasks) and in the log:
- Public network, add: an additional public subnet is delegated to the project: it is assigned from the delegated pool, and the egw gets it in its CLIENT group and routes it, as a static route, to the address next to the egw's on the project's main network - the project's igw address. The portal makes no network for it and changes nothing on the igw, which need not exist: the project's own igw (or whatever router answers at that address) takes the subnet from there. This holds however often the project is deployed again, a new main prefix included. A prefix added by an earlier release (when additional prefixes had a network of their own) keeps its OpenNebula network and its igw NIC, and the portal keeps that network's ID with the prefix. A deployment never attaches the igw to a network that only has the expected name and is not the project's: it reports it as left alone.
- Public network, delete: the route and CLIENT entry leave the egw; an older prefix's own network (in the project's group, its addresses inside the prefix) has its NICs detached and is deleted; then the prefix is released. Nothing else changes on the igw. The removal stops - Retry after a check on the platform - when the platform cannot be asked; when the recorded network is not the project's or reaches outside the prefix; or, for a prefix the project had before and was given again, when a network recorded for it is still there. The prefix stays assigned until the step succeeds.
- Main network, replace: a new main subnet is reserved at approval; the egw and igw are removed, NICs of other VMs on the old main network detached, the old network deleted and its subnet returned; then a deployment creates the new main network, egw and igw with the same settings (bandwidth, igw user, keys, ssh port and allowed networks, console password).
- Corporate network, add: a network in the project, a NIC on the cgw and the cgw configuration (interface, CLIENT group and zone); a project without a cgw gets one from a deployment. Delete: the cgw configuration, NICs of all VMs detached, the network deleted, the subnet returned.
- L2 channel: the approver gives the VLAN number (it is the VXLAN VNI too when free). On the site's management platform (it must be in execution mode) the VLAN network
s2s_<vlan>_<org>-<tenant>(802.1Q, one ethernet address, no IP) and the src gateway are created; on the project's platform the project network gets an ethernet-only address range and the dst gateway is attached through it; both gateways bridge their NIC and a VXLAN tunnel that runs inside WireGuard over cloud-s2s (link-local inner addresses). The WireGuard keys of both ends and a random pre-shared key (the tunnel's password) are made each time the gateways are made and go only into their first-boot configuration: the portal keeps them nowhere, not in Secrets either. A gateway left from an earlier run is removed and made again with the other, so both always share the new keys. Delete removes both gateways, the VLAN network and the address range. - Retry of an addition: a failed addition of a public or corporate subnet goes on with Retry only while that subnet is still assigned to the project. If the subnet was removed meanwhile (released by a − Subnet request), Retry fails at once, before any step, with "203.0.113.8/29 is no longer assigned to this project (released on …): this addition cannot be made again - ask for the prefix with a new request" (with the subnet in question). Nothing is changed on the gateways; ask for a new subnet with + Subnet.
- One change at a time: a project's network changes run one at a time. A change approved while another change of the same project runs waits and starts by itself after it, within a few minutes; Retry pressed meanwhile says so - press it again when the other change is done. A failed addition's Retry goes on only with the subnet its own approval assigned.
Router configurations are changed over ssh from the templates: the portal reads the running configuration and the interfaces, sets what the new configuration has and the router lacks, deletes only what the old configuration had and the new one does not (first-boot settings and manual additions stay), commits and saves in one session (a failed commit is discarded) and reads the configuration again to verify. Endpoints of another site are reached through that site's jump host (Sites → site settings: host, ssh port and the account whose key logs in).
Project secrets and igw configurations (project page → OpenNebula access and the igw's details; project admins, organization admins and the super admin; every look at a secret or a configuration is logged): the OpenNebula link (the platform's public URL on the FireEdge port: the site of each resource pool), the OpenNebula admin's password — Reset (a new password; OpenNebula changes too) or Keep the current password (checked by logging in to OpenNebula with it, read only) — and the igw login (address, ssh port, user, password; the portal keeps them and does not change the igw). igw configurations (VyOS igw): keep the running configuration (show configuration commands), edit a saved one (saved as a new one, linked to the one it came from), apply one. Over ssh: the portal keeps the running configuration ("before applying …"), sets what the chosen one has and the igw lacks and deletes what it does not have — never the hw-id lines nor the account the portal logs in with — and commits with commit-confirm <minutes> without saving; Confirm on the page runs confirm and save, otherwise the igw goes back to its previous configuration by itself. The timeout is asked in seconds (default 60) and rounded up to whole minutes, as VyOS counts them. By making the igw again: the configuration is checked against the igw's NICs, the igw is removed and made again from the platform's VyOS image on the same networks and addresses with the configuration on its first boot. Rolling back = applying the configuration kept before. Login: the portal's service account on an igw it made, else the project's igw login (password, through sshpass). Not tried on a real router yet.
The resource fields follow the cluster, in the three forms that set a pool's resources — a new project, Order resources on a project, Change resources of a pool: a CPU cluster: vCPU, RAM, GB and Storage, GB, typed in, with no GPU field; a GPU cluster: GPU bundles, chosen from a list; vCPU and RAM, GB shown, not typed (the bundles × the bundle's vCPU and RAM); Storage, GB typed in, or, when the cluster's bundle includes storage, shown as the bundles × that storage. A line under the fields spells the bundle out, e.g. "2 × (1 GPU NVIDIA L40S 48 GB + 16 vCPU + 120 GB RAM + 500 GB) · Bundles available: 3". The list of bundles goes up to what the cluster has free (its cards less the spare hosts' cards and the cards already allocated); when a pool's resources are changed, it goes up to the pool's bundles plus the free ones, and starts at the bundles the pool's VMs use now. While the bundle count stays, the pool keeps its vCPU and RAM as they are. The portal checks the same limits when the form is sent: more bundles than are free are refused ("Only N bundles are free on this cluster.").
Changing resources: the project's admins ask for other resources on the project's pool (vCPU and RAM, or GPU bundles on a GPU pool; storage with its VM-disk / images split; files; backups). A resource manager (or the super admin when the RM approval is off) approves it in Approval requests → Resource changes or on the project page, and can change vCPU, RAM, GPUs, storage and backups in the approval dialog first (what was asked is kept and the change is logged); the approval writes a new quota record version and, for a deployed project on an execution platform, sets the quota in OpenNebula at once. One change at a time can wait for approval; the project's admins can cancel it. The Edit button is on the Resource pools table; a waiting change is shown as a row under the current resources. The storage split is set with the pool's Storage button (images and files in GB, VM disks the rest), at once and without approval. Never below what is in use: a change cannot go below what the project uses now on the platform (the last reading, refreshed every few minutes): vCPU, RAM, GPU and VM disks of the pool; images and files over all the project's pools on the site; backups on the site. The form shows what is in use next to each field and takes nothing below it; a resource change, a backups change and an approval below the usage are refused with the list of what is in use. Removing a resource pool: a pool can be removed while the project still has VMs on it; the request then lists those VMs and asks to confirm their deletion, and the approval deletes them before the pool is removed. The site's images, files and backups must still fit in the pools that stay. When the platform has not been read, the platform itself still refuses what does not fit. A change never sets a pool's images or files below what the site stores: when they were set lower earlier, the change lifts them to what is stored (out of the new storage) and the page shows the split it will apply. A change the portal refuses says why on the page. A GPU pool is changed in bundles: a change that keeps the pool's bundle count (a storage-only change, for example) keeps its vCPU and RAM as they are, even when they differ from that many standard bundles; fewer bundles must still carry the cores and memory in use, and never fewer than the GPUs in use.
Quotas in every zone: a project's quota is set on its OpenNebula group - never on its users - in every zone where it has a resource pool: each zone keeps its own quotas, and the group (one in the federation) gets, in each zone, the vCPU, RAM, GPUs and storage of its pools there. Every approved pool order or change is followed by the quota changes; the result names each zone (e.g. "dc1: quota set … · dc2: quota set …"). When a project's last pool in a zone is removed, its quota there is reset (-1) and its ACL rules there are removed. A cluster on a platform outside the project's OpenNebula federation cannot be ordered for it (its group is not there). When a pool is ordered, changed or removed, the group's access follows it too: in each zone, every cluster with a pool has the group's cluster, datastore and host ACL rules and the zone its zone rule; a cluster without a pool has none (its quota is -1: the ACLs, not the quota, decide which clusters the group may use). Writing a project's quota to its platform is done one at a time with the portal's own quota changes for that project, so a change made meanwhile is never overwritten with an older value.
15. Settings, languages and content¶
Settings (super admin) has seven tabs; every block has its Save at the top and saves only its own fields.
| Tab | What is there |
|---|---|
| Customer | Provider - Name, E-mail (the Let's Encrypt account of the portal's certificates), Customer Portal (the portal's address and the links in e-mails), Admin Portal (the address for the platform's users; kept in the settings, not used yet), the logo (Upload logo, Remove logo), the browser icon the portal cut from it and the colours; Customers (the internal domain of the provider's hosts, adding users before they are invited, corporate email domains); Countries; Approvals. |
| Security | One row per password rule (length; how many last passwords may not be used again; small letters, capital letters, figures and special symbols at least; the allowed symbols, beside them the symbols never allowed; no part of personal data), then the igw ssh port range. |
| See Mail: the sender, DKIM and the switches, the mail sender public IPs (one per node), the DNS records with what DNS has; then the e-mail templates, one HTML template per e-mail - the selected one shows as the e-mail, Edit opens its source beside a live preview (Save, Reset to the original). | |
| Reports | The redirect e-mail above a list of the reports (name, template, day, roles, on, redirect, last sent); the selected report shows its details and sent history with Edit (name, template, day and time, roles), Turn on/off, Redirect on/off and Send now. Each report is sent with the mail template chosen for it. |
| Backups | Portal and Platforms, each with Back up now and Save at the top, the sentence "Use password ... for backup at ... and keep for ... days", the targets (a folder and, Enterprise, an S3 bucket - endpoint, bucket, region, addressing (path style or virtual-hosted), access and secret keys), and the backups under an arrow (date, time, target, size, status). Platforms also chooses the platforms (none chosen: every platform with a front-end VIP), the folders of each front-end node and Connect to Front-Ends using (each platform's SA secret, or one account secret with an ssh key for all platforms; the user needs sudo without a password, see Backup and restore). |
| Platforms | See Platforms: the Baseline (the base OS, OpenNebula and FRR releases every platform runs, Check now, the OpenNebula repository and its login) and the Installation media (the base OS installer for host deployments). |
| Portal | The node or nodes with their updates and licence in one block, then the items per page and how long history is kept (tasks, log, resource tracking), one per row, with the timers beside them. |
flowchart LR
S[Settings] --> CU["Customer<br/>provider: name, e-mail, portals, logo, colours<br/>customers, countries, approvals"]
S --> SE["Security<br/>password rules, igw ssh ports"]
S --> M["Mail<br/>sender, DKIM, switches, sender IPs, DNS records<br/>e-mail templates (HTML)"]
S --> R["Reports<br/>redirect, the reports by role"]
S --> B["Backups<br/>portal, platforms"]
S --> PL["Platforms<br/>baseline, OpenNebula repository<br/>installation media"]
S --> PO["Portal<br/>nodes, updates and licence<br/>lists, history, timers"]flowchart LR
S[Settings] --> CU["Customer<br/>provider: name, e-mail, portals, logo, colours<br/>customers, countries, approvals"]
S --> SE["Security<br/>password rules, igw ssh ports"]
S --> M["Mail<br/>sender, DKIM, switches, sender IPs, DNS records<br/>e-mail templates (HTML)"]
S --> R["Reports<br/>redirect, the reports by role"]
S --> B["Backups<br/>portal, platforms"]
S --> PL["Platforms<br/>baseline, OpenNebula repository<br/>installation media"]
S --> PO["Portal<br/>nodes, updates and licence<br/>lists, history, timers"]
Customer¶
Settings → Customer has four blocks, each with its Save at the top: Provider, Customers, Countries and Approvals.
- Provider: Name - shown in the portal's header, page titles and footer and in every e-mail (the subject starts with the name in brackets); E-mail - the Let's Encrypt account of the portal's certificates (expiry notices; a wrong address saves nothing); Customer Portal - the portal's address and the links in e-mails (its host is accepted at once as a name of the portal); Admin Portal - the address for the platform's users, kept in the settings but not used yet: its name is not on the certificate; the logo (Upload logo, Remove logo), the browser icon the portal cut from it and the colours (below).
- The certificate: the portal's certificate carries the Customer Portal's name. When the Customer Portal name changes, each portal node requests the certificate again for the new name within a few minutes (when the Let's Encrypt e-mail is set); the name must resolve publicly to the portal.
- The logo and the colours (in the Provider block): while no logo is uploaded, the portal shows Claugine's logo in its header and e-mails, and Claugine's gear as the browser icon. Upload logo (PNG or JPEG, up to 5 MB) replaces both: the portal cuts the browser icon from the logo by itself - the mark beside or above the name, on the logo's own background when it has one; a logo that is only a name gives the first letter of the provider's name on the logo's main colour. Remove logo brings Claugine's back. Colours: Claugine's (the default), From the logo, or Custom (one main colour). The portal derives the shades for light and dark pages from the chosen colour so text and links stay readable. Right after an upload the portal shows the logo's colours and asks whether to use them (Use the logo's colours / Keep the current colours); nothing changes until it is answered. An installation that had its own logo before this version, with no colours chosen, takes its colours from that logo. A logo with its own background is shown as it is on dark pages; a transparent one with dark lettering is inverted there.
- Customers: Internal domain - the DNS domain of the provider's own hosts (for example example.com). A
deployed host gets the name
<host>.<internal domain>and the domain as its DNS search domain. Required before a host is deployed; at most 100 characters. New users - whether users can be added before they are invited; the corporate e-mail domains: a list, one domain per row (+ Domain in its heading). An organization that registers with an administrator's address in one of these domains is recorded as corporate rather than cloud. Subdomains count: listingcorp.examplealso matchessales.corp.example. Staff can also mark an organization corporate when they invite it, whatever the domain. The sales rep of a corporate organization must have an e-mail of one of them. - Countries: the portal's home country, which every form offers first, and the other countries that may be chosen: all countries on the left (with a find box), the allowed ones on the right; the arrows, a double-click, Enter or Delete move a country between them. Every organization has a country, and its tax ID follows it (see Organizations).
- Time zone: the portal's time zone defaults to Asia/Almaty and is set when the portal is installed. Dates, reports and the timers follow it.
Approvals¶
- Project approval (Settings → Customer → Approvals): each approval step has a box: the resource manager (every project), the sales rep (cloud projects) and the support engineer (corporate projects). When it is ticked, every new order waits for that role: the project's or organization's own person in that role, else anyone with the role who can sign in, else a super admin. When it is cleared, nobody in that role is asked; without the resource manager's step, an order is approved as it was ordered, and if that cannot be done the step comes back to a super admin with the reason on the project. The Approvers column lists the active users of each role, and a box can be ticked only while at least one active user holds the role.
Security¶
- One row per password rule: length; how many last passwords may not be used again; small letters, capital letters, figures and special symbols at least; the allowed symbols, beside them the symbols never allowed; no part of personal data. See Passwords.
- The igw ssh port range: the range the projects' internet gateways take their ssh port from.
Mail¶
Each portal node sends the portal's mail with its own mail server, straight to the recipients' servers, under its mail name (set when the node is installed), signed with DKIM for the domain of the sender address. The portal makes the DKIM key itself; a node that already signed with a key keeps it. An installation that used a relay or an external SMTP server sends this way from this version on.
Settings → Mail: the sender address, the DKIM selector and two switches - e-mail to organization users and e-mail to platform users. Mail a switch does not allow is kept on the portal and logged (mail_held), never sent; the expiration notices of contracts and POCs follow the first switch. Send test sends a message to the signed-in user. See Email.
- Mail sender public IPs: one field per portal node - the public address its mail leaves from.
- DNS records: the records the direct sending needs (the mail names' A records, the sender domain's SPF, DKIM and DMARC, the PTR of each public address) and what DNS has now. A record is ok when it does its job, not only when it is written as the page writes it: SPF is ok when the domain has one SPF record that lets every node's address send, read in order (ip4 entries; an include, a, mx, exists or ptr before a node's address is decided cannot be checked by the page, so the record counts as another value - check it by hand); DKIM is ok when it has the same key, whatever other tags it has; DMARC is ok with any policy (the policy is the provider's choice); A and PTR when the value is among the answers. The Now column says ok or missing; for a record DNS has otherwise, it shows the value DNS has (amber), to compare with the value the page asks for.
E-mail templates: each e-mail of the portal has one template, in HTML, under one name. The e-mail's plain-text part (for mail programs that show text only) is made from the HTML when the e-mail is sent: paragraphs, lists as "- " lines, tables as rows of cells, links with their address. The password reset's subject line is the one text template. The list shows each template's Name and Description, with the filter on its title row; select a template to see the e-mail as a page (its variables shown as ‹name›). Edit opens the template with its name on the dialog's title row, then Save and Cancel; the source, with the template tags and variables coloured, and the e-mail redrawn as you type are below (a template that does not compile is not saved). Reset to the original goes back to the portal's own. The name is yours to change: the portal sends every e-mail by the template's fixed number, so renaming a template changes nothing that is sent, and a report keeps its template when it is renamed. Two templates cannot have the same name. A template edited as text before this version is carried over as HTML. Every e-mail is sent in the provider's frame. The texts of the reports are among the e-mail templates.
Reports¶
The portal e-mails four weekly reports: Super admin, Support, Sales (manager and KAM) and Resources.
- Settings → Reports: the redirect e-mail above a list of the reports (name, template, day, roles, on, redirect, last sent); the selected report shows its details and sent history (90 days) with Edit (name, template, day and time, roles), Turn on/off, Redirect on/off and Send now. Each report is sent with the mail template chosen for it. A report's Template is chosen from the report e-mails by their names. The redirect e-mail is the one address the reports marked Redirect go to instead of their recipients.
- A report reaches every holder of its roles who has an e-mail address, invited or not; blocked and archived users are left out. A support engineer, a manager and a KAM get their own organizations only; the super admin and the resource manager get everything.
- Every report starts with its Totals row: tiles with the count and what was +new, updated and −archived in the period. Then, by report:
| Report | Totals tiles | Tables after Totals |
|---|---|---|
| Super admin | platforms, clusters, hosts, organizations, projects, platform users, organization users | Platforms; Clusters (every kind, physical view); Datastores (OpenNebula's numbers); IP subnets; Backups (every run: type, date, status, size); Portal updates (date, version, standby node, active node; a pair's update is one row per release, with the state of each node, the standby first, also when it stopped half way) |
| Support | clusters, organizations, projects | Platforms; Clusters (every kind, physical view); Datastores (all, OpenNebula's numbers); IP subnets; Projects (quota / in use) |
| Sales | organizations, projects, organization users | Projects (start and review dates, quotas) |
| Resources | shared platforms, shared clusters, organizations, projects, organization users | Shared clusters (logical view); Datastores of the shared platforms (the projects' quotas of the usable size); IP subnets; Projects (quota / in use) |
A table the report has is always there; when it has no rows it says "None.". A datastore OpenNebula has not reported yet is listed as "not read yet". Organizations, projects and IP subnets link to the portal. The meters are coloured by the same rules as Analysis → Summary (see Platforms and the colours of capacity).
Backups¶
- Portal and Platforms, each with Back up now and Save at the top, the sentence "Use password ... for backup at ... and keep for ... days", the targets (a folder and, Enterprise, an S3 bucket - endpoint, bucket, region, addressing: "path style" or "virtual-hosted", access and secret keys), and the backups under an arrow (date, time, target, size, status). Platforms also chooses the platforms (none chosen: every platform with a front-end VIP), the folders of each front-end node and Connect to Front-Ends using: the account the portal logs in to each front-end node with - by default each platform's SA secret (the same login the portal uses for every host and VM), or one account secret (with an ssh key) chosen for all platforms. The user needs sudo without a password on the front ends.
- The S3 row of each block (Portal, Platforms) has an optional Region. Empty, the portal takes the region from the
endpoint when the endpoint names one in the form s3.
. - e.g. Backblaze B2's s3.eu-central-003.backblazeb2.com gives eu-central-003, AWS's s3.eu-west-2.amazonaws.com gives eu-west-2 - and otherwise signs for us-east-1, which most self-hosted S3 stores (MinIO, Ceph) accept. Set it only for a store that wants another region name. - Both go into one folder, side by side:
claugine_<portal address>_<YYYYmmdd-HHMM>andplatform_<platform>_<address>_<YYYYmmdd-HHMM>(a backup made in the same minute as the one before it can be named with its seconds as well; in a bucket an existing folder is never written over, and such a backup may be reported as a clash instead, with no copy in the bucket); each kind is listed and pruned only among its own. A backup file larger than 64 MiB is uploaded to the bucket in 64 MiB parts (multipart upload); a failed upload is aborted. The recommended bucket policy allows s3:PutObject, s3:GetObject and s3:AbortMultipartUpload under the prefix (aborting a failed upload deletes no backup). Everything about them is in Backup and restore.
Platforms¶
- Baseline: the releases every platform runs: the base OS (an Ubuntu LTS release, such as 26.04), OpenNebula (major.minor, such as 7.4) and FRR (major.minor). The portal reads every platform's front ends and hosts once a day (the timer "Platform baseline") and on Check now: the OS and OpenNebula of each node, read on the node itself; FRR is shown, never counted. A platform with a node on another OS or OpenNebula release shows upgrade required; a node that was found out of date stays so until a read shows it on the baseline or OpenNebula no longer lists it.
- After a change of the OS or OpenNebula value, every platform shows not checked against these values yet until a complete check (the platform and every node read) has run against the new values; a check asked for while one runs runs again when it ends. Changing FRR alone needs no new check.
- While a platform is out of baseline or not checked against the current values, no host is deployed and no platform is created; the refusal names the platforms. Between changes of the values, a node that cannot be read blocks nothing; it is shown as not read.
- The same block holds an OpenNebula repository (an https address, by default the OpenNebula download site) and, when the repository needs one, its login, chosen from the account secrets. New hosts install OpenNebula from it.
- Installation media: the portal keeps the installer of the base OS for host deployments: once a day and before every deployment it looks for the newest point release of the base OS on releases.ubuntu.com, checks its signature and checksum, and rebuilds it to install with no questions. The current and the previous installer are kept, and any installer a running or failed deployment still uses. With two portal nodes, the other node is given the same installer and checked until it has it; each site's hosts take the installer from the portal node of their own site. The block shows the release, when it was rebuilt, its SHA-256, where it is served for each site, the other node's copy, and the hosts whose iLO address is not an IPv4 address (they cannot be served). The installer is served only to the iLO networks of the site's hosts.
Portal¶
Settings → Portal shows the portal's node(s) with their updates and licence in one block, the actions on top:
- Update OS: the OS updates, one node at a time on two nodes (see Software delivery and updates);
- Update portal: checks in and installs the newest release offered to this installation, one node at a time on two nodes, after a backup; a release older than the one the portal runs is never installed; that backup keeps the data from before the update, restored on the current release (see Software delivery and updates);
- Change encryption key: the key to the stored secrets (a backup, the new key on every node, every secret encrypted again, the old key removed, a backup with the new key; the portal restarts twice, for seconds). Change encryption key works again (in v01.00.37 and v01.00.38 it failed: it restarted the removed test version). A failed restart puts the keys back as they were;
- Switch over (two nodes): the standby takes over; refused while tasks run or when the nodes run different versions.
Update OS, Update portal, the edition upgrade, Change encryption key, Add second node and Replace node are maintenance. While one of these runs, the portal starts no new work: a provisioning, a network change, a deletion, a configuration check, an igw apply, an archival or a task asked for meanwhile is refused with "Maintenance is running: try again when it ends" (the API answers 503). The operation first stops new work, then checks that nothing is running (else it does not start: wait until the work ends), and lets work start again when it has finished, done or failed. On a pair, Switch over and both halves of Update portal and Update OS stop new work the same way before they check that nothing runs, and keep it stopped through the change of roles; the node that takes over starts no new work either until the operation has ended. A platform's Make consistent (Infrastructure → the site's Platforms) counts as running work: an update waits for it (it does not start while a repair runs), and Make consistent is refused while maintenance runs. A repair cut off by a restart is shown as interrupted - run Make consistent again.
The table lists each node: its site and address, role (active, standby, maintenance), release, database (writable or read-only, with how many seconds a standby is behind) and its last operation. The steps of every operation are in Administration → Logs (the Logs link).
On two nodes, one operation at a time runs on a node, and one cut off by a restart is shown as interrupted; a switch over that stops half way never leaves two active nodes; an update that stopped half way is finished by Update portal; each node knows the other's certificate, and a node reinstalled with a new certificate is paired again with Replace node. See Safeguards of the pair.
Below, Updates: Automatic updates (install new releases by themselves; the stable or all-releases stream), the licence (its state, the installation ID — the portal's public host name — and the last check-in; see Licence) and New build (a newer release offered, or up to date).
- HAProxy on the portal VMs serves the portal's name and the OpenNebula front ends. Each portal VM applies a
change by itself within 5 minutes; a name that is new gets its Let's Encrypt certificate then (when the Let's
Encrypt e-mail is set in Settings → Customer → Provider). When HAProxy is set up again right after an update, the
update may still be installing packages; the set-up then waits for the package manager up to 10 minutes instead of
failing. Break-glass only (the portal's UI is down):
sudo claugine-lb apply, orsudo claugine-lb certificatefor new names, on each portal VM. -
OpenNebula front ends: HAProxy on the portal VMs serves each shared platform's front end under the platform's public FQDN, on 443:
Service Address Open to FireEdge https://one-dc1.example.com/fireedge/sunstone(the name alone opens it too)anywhere XML-RPC API https://one-dc1.example.com/RPC2anywhere OneFlow https://one-dc1.example.com/oneflowanywhere Each asks for the OpenNebula login itself; they work when the name is proxied by Cloudflare too.
Each site's front end is its shared platform that has a Public FQDN (the first by name when there are several); the name is set in Infrastructure → the site → Platforms → Edit. Nothing about the front ends is set in Settings. A project's OpenNebula addresses — the link on its page and the addresses in the credentials and project-ready e-mails — are these three. Clients:
ONE_XMLRPC=https://one-dc1.example.com/RPC2,ONEFLOW_URL=https://one-dc1.example.com/oneflow. The gRPC API is not published: OpenNebula's own gRPC clients (the CLI and the libraries) connect without TLS. OneGate is never published by HAProxy (/onegateon the front end's public name is FireEdge's like any other path): VMs reach it through their own host athttp://169.254.16.9:5030(OpenNebula's transparent proxy), or at the front end's VIP when the platform's OneGate through hosts is cleared (see OneGate). Publishing needs no change on the front end for FireEdge and the XML-RPC API: FireEdge already serves under/fireedge, and oned's XML-RPC path is/RPC2. -
Nodes and Updates: see Portal architecture and Software delivery and updates.
- Add second node (an Enterprise installation that runs one node): the page shows this node's key, to add to
the other VM's user; enter that VM as
user@address, then Check only (the checks, nothing changed) or Add second node; each step shows as it runs. Replace node, on the same card, when the other node stops answering: give a new, clean VM, and the active node installs the portal there, copies the database to it and makes it the standby. See Setting up the second node. - Releases: bundles from the control plane, and the latest release. The nodes take every release as a bundle from the licensing portal; there is no source repository to set (a pair's standby still gets each release first).
Licence¶
Settings → Portal shows the licence of this installation:
- the state: active, grace, expired (after the 60 days of grace: updates stop, the portal keeps running), not yet in force (the licence starts later), conflict, retired, or awaiting check-in before the first check-in has answered;
- the licensing portal it checks in with, and the installation id: the portal's name (the host of its site URL), which binds this installation to your licence. Installing again or restoring under the same name is the same installation;
- the release stream, and the time of the last check-in;
- for a Standard installation, what is pinned: stable releases only, installed in one window a week.
Updates run while the licence is active or in grace. A release my.claugine.com marks critical is installed whatever the licence state - also when it has expired or was revoked - as any update: by automatic updates in their window, or at once with Update portal, one node at a time on two nodes. Settings → Portal marks it "critical" and, when updates are otherwise stopped, says it is installed all the same. Releases that are not critical stay stopped. The installer and a restore are not changed by this.
An administrator reads this block; nothing on it can be edited. Two states carry an explanation on screen:
- Grace: "The licence has lapsed; updates continue until date, then stop. The portal keeps running."
- Retired: this installation was released from its licence and receives no more updates; decommission the node. A portal under another name is a new installation.
Lists, Keep history and Timers¶
Below the nodes block, Settings → Portal has the items per page and how long history is kept (tasks, log, resource tracking, in days: older entries are removed every night; projects and organizations never), one per row, with the timers beside them.
Timers: each with its period in minutes, the last run, its state (ok, in progress, error) with a short result, and a Refresh button (locked until it is over):
| Timer | Default | What it does |
|---|---|---|
| Portal | 10 min | The portal's components: the database, both nodes and the replication, the nodes' mail servers, the last backup. |
| Platforms | 30 min | The platforms' endpoints and service images; the projects read from the platforms (groups, quotas, users, VMs, internet gateways — their VM, size and NICs); every platform's datastores, hosts and VM counts and the platforms' own VMs (gateways, core services) — the numbers of the dashboard. |
| IPMI | 240 min | The hosts' health from their IPMI. |
| Platform baseline | daily | Every platform's front ends and hosts read against Settings → Platforms → Baseline (the OS and OpenNebula of each node). |
| VyOS | 480 min | VyOS routers compared with their templates: route reflectors, site egw/igw, project egw, cgw and L2 gateways. |
| Registrations | daily, from 09:00 | Deletes the registration requests not confirmed in time and the declined join requests (the log keeps them). |
| Reports | 5 min | Sends the e-mail reports that are due (Settings → Reports); on the active node only. |
The dashboard says which timer its numbers come from and when it last ran (clusters, storage, VMs); the super admin has a link to the timers.
flowchart LR
T[Platforms timer<br/>every 30 min] --> P1[Project platforms:<br/>projects, quotas, VMs, internet gateways]
T --> P2[Every platform:<br/>datastores, hosts, VM counts]
T --> P3[Every platform's own VMs:<br/>gateways, core services]
P1 --> TP[Project pages]
P2 --> DB[Dashboard: clusters and storage]
P3 --> GW[Gateways and core services]flowchart LR
T[Platforms timer<br/>every 30 min] --> P1[Project platforms:<br/>projects, quotas, VMs, internet gateways]
T --> P2[Every platform:<br/>datastores, hosts, VM counts]
T --> P3[Every platform's own VMs:<br/>gateways, core services]
P1 --> TP[Project pages]
P2 --> DB[Dashboard: clusters and storage]
P3 --> GW[Gateways and core services]
Status of the core services¶
A site page (Infrastructure) refreshes the statuses of its core services by itself only while one of the site's tasks runs — every 10 seconds; otherwise with Refresh status.
Languages and content¶
- The version: platform users see the portal's version at the right of the bottom line of every page.
- Languages: the portal is available in English, Russian and Kazakh — switch in the top bar. The list is not a setting: English is always present, and each other language appears because its translations are part of the installed version.
- Content: news, announcements and incident banners with an optional end time, and public service status with 30-day uptime on the main page.
16. Email¶
Each portal node sends the portal's mail with its own mail server, straight to the recipients' servers, under its mail name (set when the node is installed), signed with DKIM for the domain of the sender address. The portal makes the DKIM key itself; a node that already signed with a key keeps it. An installation that used a relay or an external SMTP server sends this way from this version on. The switches, the test message and the DNS records to add are in Settings → Mail (see Mail).
Every e-mail of the portal is sent in HTML (with a plain-text part made from it), in the provider's frame with its logo carried inside the message (shown without loading pictures from the internet); addresses and links keep their own text. Names of people are e-mail links; the organization's name is in bold. Who gets which e-mail is listed in Ordering and provisioning.
On provisioning, the project admin gets the access details, with the platform user who provisioned the project in copy; every organization admin who is not the project admin gets the same e-mail of their own. Only when the project admin's e-mail went out are the project's platform users told. "Resend credentials" sends it to the project admin only. Every e-mail has its template (Settings → Mail → e-mail templates).
17. Security and data¶
- Secrets are never in code: the Django key, the database connection and the encryption key are in portal.env; every other secret (platform passwords, keys, the mail and backup secrets) is encrypted in the database.
- Every user, staff and customer, must set up two-factor authentication with an authenticator app (TOTP) before doing anything else. After the password, the second step of the sign-in takes either the authenticator app's code or a passkey. A person adds passkeys on their profile page: Touch ID, Face ID, Windows Hello, a phone or a security key. The authenticator app stays, as the way back when a device with the passkey is lost. A passkey works only on the portal's own address. Each code is accepted once; after a wrong code or a failed passkey the next attempt waits a moment, longer after each further failure, and after five failed attempts the sign-in starts again.
- The technical admin site (
/django-admin/) has no sign-in of its own: it sends to the portal's sign-in, with its second factor, and opens for a staff session made there. The technical administration pages show people and invitations but do not change them: a person's roles, e-mail, sign-in and two-factor authentication are changed only on the portal's own pages. - Every significant action is written to the audit log (who, what, object, details).
- Photos are served only to users allowed to see the person.
18. Operations¶
- Install, update, back up, restore: see Running the portal.
- Management commands (
sudo claugine-manage …):bootstrap_superadmin(one-time super admin link),collect_platforms(collect now), plus standard Django commands; as root,sudo claugine-manage-root ha_agent …(the HA agent). - Logs: the application's own messages in the systemd journal (
journalctl -u claugine-app); Apache's access and error logs under/var/log/apache2; mail in/var/log/mail.log; the timers and the HA agent in the systemd journal.
19. Roadmap¶
- Provisioning engine: desired vs actual state, reconciler, procedures, collectors; test projects.
- The task API (customers already ask for resource, network and L2 changes on the project page).
- Automatic lifecycle transitions and personal-data deletion on request.
- Dedicated clusters.
- Production platform endpoints.