A geographic map of your FortiGates, and the topology graph behind each site.

Gated by deviceMap:read — the whole mount, not just the page.

Drawing regions needs mapRegions.

---

OpenStreetMap tiles with Leaflet. Pins are FortiGates, coloured by monitor health and clustered at low zoom.

Colour
the automation's own severity	down — red is what critical looks like, not what Down means
amber	warning
slate blue	dependency-suppressed — suppression outranks the device's own state
purple	maintenance
green	up

A cluster counts suppressed children as dep-down; a non-suppressed down or degraded child still rolls the cluster up red or amber.

Click a pin for the site; click through for its topology.

Clusters. Hover a numbered circle to list the FortiGates inside it, worst status first. Click a name in that list to open its topology. Click the circle itself to zoom in until the gates separate. Gates that share a location, or sit so close together that no zoom can separate them, fan out in a ring instead. Each one gets a line back to a dot that marks the real location. Click the map or zoom to fold them back. From zoom 11 a cluster forms only where pins would overlap.

The sun / moon button in the map's top-left corner, above the zoom buttons, switches the basemap between light and dark tiles independently of the app theme. It is remembered per user in this browser; until you touch it, the map follows the app theme's family (Morning and Noon light, Nightfall dark).

Three tiers, highest first — see Fortinet discovery. A manual coordinate source pins them against discovery.

Coordinates are validated as a pair; a half-valid tier falls through rather than mixing values.

A reverse proxy in front of Polaris must not add its own Referrer-Policy. OpenStreetMap blocks referer-less tile requests, so a stripped Referer turns every tile into "Access blocked". This is the single most common map problem.

Tiles come from one host, and the OSM usage policy applies to your install.

---

Polygons drawn on the map that auto-tag the gates they enclose. A region is also a registry tag — region:<name> in a locked category.

Regions nest, and the nesting is load-bearing. Alert routing can address "the device's own innermost region" (L1) or "the division containing it" (L2) — see Actions.

The Regions button in the map's top-right corner, ungated beyond viewing the map — how an install's regions nest is not privileged information. Clicking it opens the Regions panel, which lists the regions as their nesting tree — each region indented under the region that contains it, with its colour and level; the panel's × folds it back into the button.

Open, it paints the polygons read-only, with a permanent label only on the outermost ring of a nested set (fill opacity is held low because nesting stacks it), and hovering renders the containment tree, each row badged with its derived level. Smaller regions always draw on top of the regions around them (in edit mode too), so an L1 inside an L2 is the one you click.

In edit mode, drag a region's row in the panel onto another row to nest it there. Only a region whose polygon contains it on the map accepts the drop — this chooses between real containers, it does not override the map. The case it exists for: a sub-region inside a state inside a division can be nested directly under the division, and the levels re-compute (the state, left with nothing inside it, drops to L1). A refused drop says why in the status line. The change saves immediately, like a rename, and is recorded in Events (region.nesting_changed, plus region.levels_shifted for the other regions it moved). A region nested by hand shows its name in italics; drop it back on its smallest container to return it to automatic nesting. If a later reshape means the chosen region no longer contains it, it goes back under its smallest container until the shape is restored. See Rule 121.

Needs mapRegions: the pencil left of the Regions panel's title. Draw, edit and delete polygons. While you edit, the panel's header holds the controls:

Control
✓ (where the pencil was)	save every reshaped region and leave edit mode
Undo / Redo (or Ctrl+Z / Ctrl+Y)	step back and forward through vertex drags, added points and deleted points
×	discard unsaved shape changes and leave edit mode; it asks first when there are any

Undo covers only shape changes, which stay local until you save. Drawing, renaming, recolouring and deleting a region are saved the moment you do them, so undo cannot reverse them.

The two region layers hand off rather than stack: entering edit mode hides the read-only overlay and disables the Regions button; leaving restores it and drops the memoised payload, since an edit can have changed every level in the tree.

The columns holding region assignments on users, roles and group mappings hold bare names with no foreign key. So:

• A rename carries them with it. Leaving them behind would revoke every scoped operator's region in silence — tag present, matching no region, and every name-resolving consumer quietly reaching nobody.
• A delete never strips them. There is no new name to move an assignment to, so the assignment survives and a warning Event names who now holds a dangling one.

Renaming or deleting a region records the retired name, and a sweep strips region:<name> from every asset and network still carrying it — both halves required (rule 54).

Both halves are needed because the two existing strip paths are blind to this case by design: the reconcile removes only pairs it recorded, keyed by region id, which a rename does not change and a delete drops outright.

The obvious fix — strip every region: tag matching no region — is the one thing this must not do, because manual attachments survive every reconciler forever and an operator could hand-apply region:Narnia to a printer.

A name that is live again is reclaimed, not stripped.

Why this matters beyond tidiness: an orphaned region tag makes an automation's level-scoped routing abstain entirely (rule 58), rather than promote the container. Switches tagged with a misspelled retired name resolve every down alert's L1 to the division, mailing the division's recipients while the people scoped to the site are never reached — and the automation looks "working" by every check the UI can offer.

A read-only strip in the toolbar: your own effective region scope as coloured pills, each tooltipped with where that scope comes from — your account, your role, or an IdP group. An empty scope is stated as "all regions", since empty means unrestricted.

It answers "which regions am I assigned to?" — which the Users page shows only to whoever administers users — and is deliberately not a filter: the map already shows exactly the sites you may read.

---

Opens from a site. A Cytoscape graph of the gates, switches and APs at that site, with a column solver for layout and a dagre fallback. The gate sits at the left and each switch one column further right than its uplink; a switch's APs stack beneath it in the column to its right. A link that leaves a switch for a child switch on another row turns up or down right next to the switch and runs along the child's row, so it never shares a line with the switch's straight link to the next switch along, and lines travel the gutters instead of cutting across other devices. Drag anything to override the layout; the positions are saved per site.

Port labels (port8 ↔ port1) on a link with bends sit near its ends: with one bend, halfway along the part after the bend; with two or more, halfway along the first part and again halfway along the last. A label is never put where several links run along the same line, since it could belong to any of them: it moves to a stretch only its own link uses, and a link that shares its whole route shows no label.

On the light themes (morning and noon) the links and location boxes are drawn in darker shades of the same colours, so they stay readable on the pale background; the legend matches the theme you are using.

With Locations on, the details pane lists the switches and APs in the same nesting as the location boxes (area › building › floor › room › junction box), under a heading per group, with devices that have no location last.

Control
Search	find a node
Show full	the whole graph rather than the narrowed view
Refresh	re-read
Save layout	deviceMap:write — shared across operators; readers fall back to browser storage
Reset layout	a menu: Reset to last save (where a restore point exists) or Reset to baseline — the column solver's layout
Snap to grid	per user; lands drags on the solver's lattice and re-snaps everything on enable
Base / area / building / floor chips	top left, only when a device carries a b: or f: location code. Base is the whole site; under it are the areas (plus any building with no area), or the buildings when no area is tagged. Picking a chip shows its devices and lists the next level beneath it: an area's buildings, then a building's floors. Each view keeps its own saved layout
Fullscreen
Export	screenshot, PDF or Visio — see below

The legend sits at the graph's bottom left as a small button; click it to open the legend, and its × folds it back. The edge between the graph and the details pane hides the pane: hover the edge and click the arrow, and the arrow at the right edge brings the pane back. Both remember your choice in this browser. The Device Map's own status legend, the Application Map's and the Status Map widget's work the same way.

Save is an explicit checkpoint that writes the live layout and the restore point in one statement, so a checkpoint can never be of a layout the server has not stored.

The Export button at the top right of the topology header opens a menu. Each choice exports the graph as it is on screen, in its current layout, including a narrowed endpoint-search view.

• Copy screenshot (map + details): the graph and the right-hand details panel as one image. Clipboard first, download as fallback. The small camera inside the graph area still copies the graph alone.
• Export PDF…: the graph drawn as vectors on white paper whatever theme you are using, so it prints sharp and its text is searchable. Choose the paper size, orientation and minimum text size (7 pt by default): nothing on the printed graph is smaller than that. A site that fits one sheet at that size prints on one; a bigger one is split into overlapping sections after an overview page with the section grid, and empty sections are skipped. Every page carries the legend (health colours, link types, location groups) in a box at the bottom left. Characters the PDF fonts cannot print are simplified: ↔ prints as <->.
• Export to Visio (.vsdx): one Visio page sized to the graph, every device, location box and link a separate editable shape, health colours and link styles kept. Links are drawn lines, not glued connectors (moving a device in Visio does not drag its links), and text uses Visio's default font. The title and summary line sit above the graph, and the legend sits in a box at its bottom left, as on the PDF.

The Application Map has the same menu (Application Map).

A standby member is dropped from the graph. It would otherwise hang off the same switches as the primary and duplicate the cluster. The active member represents it; the asset's own dependency panel shows the second box as a tagged sibling row.

The dependency graph prefers the most physical signal available — mesh > interface > LLDP > controller — and each kept edge records which put it there.

Two overrides worth knowing: a mesh-leaf AP depends on its root AP rather than its controller, and a FortiLink switch bridged behind a FortiAP depends on the AP.

Each switch's info panel lists the endpoints learned on its ports (a count and the 25 most recently seen), and the search box finds an endpoint and pulses its switch. A device is placed under the switch the dependency tree resolved for it — so when several sites name a switch "IDF-1", each site's panel shows only its own devices (rule 91). A device whose switch name is shared and that nothing could settle is left off the panel rather than shown under the wrong site.

---

A different surface: the Site Map dashboard widget is a geographic map of monitored sites with status dots and live weather radar (proxied server-side, with a direct-CDN fallback).

The Device Map widget is the gate-level map with click-through to topology.

On both widgets the sun / moon basemap button sits in the map's top-left corner, above the zoom buttons, as on this page. The Site Map's °F, Radar and Loop buttons, and fullscreen, are at the top right; a button that is on looks pressed. Like the rest of the map's buttons and its legend, they follow the app theme.