Integrations

Integrations

The page where you connect Polaris to the systems that know things.

Gated by integrations. Reads at read; creating, editing and running discovery at write.

No integration is guaranteed and every one is absent by default — discovery is something you turn on, not something you turn off. An install with none still works on hand-entered assets and address space.

But this page is the point of the product: pulling the systems you already run into one dashboard, where a device known to four of them is one record carrying four sources rather than four rows to reconcile by hand.

---

The eleven types

Type	Reads	Produces	Page
FortiManager	many FortiGates via FMG	assets, networks, reservations, VIPs	Fortinet
FortiGate (standalone)	one device over REST	same	Fortinet
Entra ID / Intune	Microsoft Graph	assets	Directory
Active Directory	LDAP / LDAPS	assets	Directory
Windows Server	WinRM DHCP	networks (from DHCP scopes)	Windows Server
VMware vCenter	vSphere REST + SOAP	assets, datastores	vCenter
Azure Arc	Azure Resource Manager	assets	Azure Arc
Unraid	Unraid GraphQL API (7.2+)	host, VMs, containers	Unraid
TrueNAS SCALE	TrueNAS JSON-RPC WebSocket API (25.04+)	host, VMs, Apps	TrueNAS SCALE
Proxmox VE	Proxmox VE REST API (9.x), read-only	cluster nodes, VMs, LXC containers	Proxmox VE
Generic API	any REST API returning JSON, mapped field by field	assets	Generic API

Plus one integration that discovers nothing: AI Assistant — the model server behind the AI assistant, either one you run or an Azure AI Foundry deployment.

The page has a second tab, Polaris Agents — the Polaris Agent build, SSH deployment, and service/process discovery rules — which is not an integration row.

Network Discovery — saved active scans of IP ranges — lives on the Assets page under + Add Asset(s).

---

Adding one

+ Add Integration → pick the type → fill in the modal.

The modal is tabbed, and the tab set depends on the type:

Type	Tabs, in order
FortiManager / FortiGate	General → Filters → Monitoring → DHCP Push → Quarantine Push → Description Sync → SD-WAN → Geographic Location
Active Directory	General → Monitoring → Directory
Entra ID / Intune	General → Monitoring → Directory → Script Publishing
Azure Arc	General → Monitoring → Script Publishing
Windows Server, vCenter, Unraid, TrueNAS SCALE, Proxmox VE	General → Monitoring
Generic API	General → Records & Mapping → Preview (its assets are monitored one by one, like manually added ones)
AI Assistant	one untabbed form — Model Server, then Assistant Behaviour (AI assistant)

Outside the Fortinet pair, the connection settings and the filters live on the General tab. The FortiManager and standalone FortiGate layouts are deliberately identical, diverging only where the two integrations genuinely differ.

Fields every type has

Field
host	absent on Entra and Arc, whose endpoints are fixed
port
verifySsl / verifyTls	on by default. Windows Server has Use SSL instead, and Entra and Arc talk only to Microsoft's fixed endpoints
verboseLogging	always the last element on the General tab

Plus, on the integration row itself:

name
enabled
autoDiscover	on by default
pollInterval	hours, 1–24 — 12 for most types, 4 for Windows Server, 1 for Unraid, TrueNAS and Proxmox VE

Test Connection

Every modal has one, and it records the result on the row. Test before you save, and again after changing a credential.

The button knows which fields it needs per type, and on the edit path it drops blank secrets — so "leave blank to keep the current secret" works. That holds only while the host, port, token URL and connection type are unchanged: change one of them and the secret must be entered again, because a saved secret is never sent to a new address (rule 111).

---

Running discovery

Either the row's Discover button, or the scheduler on pollInterval.

Both wait for a successful Test Connection. Until the row has one, the Discover button is disabled and the details panel reads Disabled until a successful connection test — and a row whose last test failed stays skipped by the scheduler until a manual test passes again.

The Discovery Activity dashboard widget shows in-flight runs with per-run progress and amber telemetry for slow ones; the sidebar carries a status panel for the same thing.

Both render · N offline and · N unread as separate parts, and no surface may sum them (rule 53):

offline	routine — a staged gate awaiting deployment sits offline for weeks, and Polaris reads its cached configuration on purpose
unread	the device was never reached, and is named in a warning Event

See How discovery works for the model underneath all of this — the multi-source projection, presence rules, post-sync passes, and how each type decides something has gone away.

---

Monitoring settings

Every integration carries a Monitoring tab: the integration tier of the polling-method hierarchy, split into per-class blocks.

Type	Classes
FortiManager / FortiGate	FortiGate · FortiSwitch · FortiAP
Entra / AD / Windows Server / Arc	Workstations · Servers (Arc adds Kubernetes)
vCenter	VMs · ESXi hosts
Unraid / TrueNAS	Host · Virtual Machines · Containers (Apps on TrueNAS)
Proxmox VE	Nodes · Virtual Machines · Containers

Each block carries addAsMonitored, per-stream polling methods and credentials, and — on the classes that support it — agent auto-deploy and interface/storage auto-monitor. The AD, Entra, Arc and vCenter VM blocks carry the full set; Windows Server, ESXi hosts, Kubernetes clusters and the Unraid / TrueNAS / Proxmox VE classes carry addAsMonitored and streams only.

AD, Entra, Arc and vCenter also carry Verify network presence after discovery (on by default) at the top of the tab — see Post-sync passes.

Cadence and retention live on the same tab.

Turning on addAsMonitored starts polling a class of devices. At fleet scale that is a real load decision, and it is also the gate that decides whether automations may fire about them at all.

---

Push toggles

Only the Fortinet types write to devices, and every push toggle is off by default:

Toggle
DHCP Push	manual reservations are written to the gate, verified by read-back
Quarantine Push	MAC address-group entries
Auto-reserve Fortinet infrastructure	requires DHCP Push
Adopt discovered MAC	requires DHCP Push
Description Sync	writes Polaris descriptions to devices
Coordinate write-back	writes geocoded coordinates

Two Azure-side write capabilities are also off by default and gated separately: Intune script publishing and Arc Run Command.

Unraid and TrueNAS add one more: starting, stopping, restarting and updating a VM or container from its asset, by a user with Assets Read-Write and only when the integration's API key is allowed to (rule 94; see Unraid and TrueNAS SCALE). Proxmox VE is monitored read-only (Proxmox VE).

Everything else Polaris does is a read.

---

Diagnosing an integration

Three tools, in order of how much they cost:

1. Test Connection — is the credential still good?
2. The Query API button — proxies a raw read to the upstream system. It is the self-service way to answer "why didn't device X get discovered?". Every type has one except Windows Server.
3. Verbose logging — the bottom of the General tab. The next discovery cycle and every monitor job for this integration's assets emit step-by-step logs at info level. High volume; turn it off when done.

Then read the run's Events, filtered to warning and error.

---

A sensible order for a new install

1. Server Settings → Credentials first. Add the SNMP, SSH, WinRM and REST credentials before the integrations that reference them.
2. One integration, tested, discovered once, and its results reviewed — before adding a second. Two integrations discovering the same devices is normal and correct, but it is easier to read one source's output first.
3. Leave addAsMonitored off until you have seen what a class actually contains.
4. Leave every push toggle off until the read side is right. A push is a write to your network.
5. Then automations, and give the seeded ones real recipients.