devices(ubiquiti): UCG-Ultra and UDM-Pro accept no PoE, so their copper ports are not PDs #2124

Merged
jaylouisw merged 2 commits from catalogue/1119-0852 into main 2026-11-19 10:51:39 +00:00
Owner

Refs #130.

isPoePd() matches unifi in model + ' ' + label, so every RJ45 socket on a UniFi gateway
that does not declare its own poeIn was filled in as a 12.95 W 802.3af powered device. Two
gateway rows in devices/ubiquiti/ were still in that state; the UXG family was cleared last
pass (#2121).

row ports drawing phantom load before after
devices/ubiquiti/UCG-Ultra.json wan1, port1-port4 64.75 W 0 W
devices/ubiquiti/UDM-Pro.json wan1, port1-port8 116.55 W 0 W

Measured through the engine, not by hand: new Device('x', spec) and summing
interfaces.hardware.poeIn on origin/main versus this branch.

What the vendor says

UCG-Ultra — https://techspecs.ui.com/unifi/cloud-gateways/ucg-ultra (HTTP 200, 657,109 B):
Port Layout (4) 1 GbE & (1) 2.5 GbE, Max. WAN Port Count 4, Max. Power Consumption 6.2 W,
Power Method USB type C (5V DC/3A), Power Supply USB type C (5V DC/3A) (Adapter included).
Zero occurrences of PoE or 802.3 in the document.
https://store.ui.com/us/en/products/ucg-ultra (200, 550,854 B): PoE Support flag Empty,
PoE Budget flag Empty, Max. PoE Wattage per Port by PSE flag Empty.

UDM-Pro — https://techspecs.ui.com/unifi/cloud-gateways/udm-pro (200, 676,527 B):
Port Layout (9) 1 GbE & (2) 10G SFP+, Max. Power Consumption 33 W, Power Method
(1) Universal AC input, 100-240V AC, 50/60 Hz + (1) USP-RPS DC input, Power Supply
AC/DC, internal, 50W. Zero occurrences of PoE or 802.3.
https://store.ui.com/us/en/products/udm-pro (200, 600,589 B): PoE Support, PoE Budget and
Max. PoE Wattage per Port by PSE all flag Empty.

So the watts were not merely unverified, they were impossible: a 6.2 W USB-C console and a 50 W
AC/DC console, neither with a PSE, each "receiving" more power over copper than it can consume.

Traps recorded

  • The store page is not a single product's spec sheet. ucg-ultra's payload contains
    Power Method = PoE / IEEE 802.3af and copy for a "PoE Integrated AC Adapter" and a
    "PoE-powered 5G RedCap antenna". Those are cross-sell SpecificationEntity blocks for other
    SKUs; the entries that belong to this product are the ones carrying the Empty flags above.
    Reading the first PoE string on the page would have made this console a PD for real.
  • No UCG-Ultra datasheet exists. https://dl.ui.com/ds/ucg-ultra.pdf and
    .../ucg-ultra_ds.pdf both answer 302 -> https://techspecs.ui.com - the same silent
    fallback already recorded for uxg-fiber, not a redirect to a PDF.
  • devices/ubiquiti/UDM-Pro.json carried two sourceUrl keys; JSON kept the last
    (https://dl.ui.com/ds/udm-pro.pdf), so the row was not uncited, but the duplicate is gone.

Gates

node scripts/syntax-check.js            Syntax-checked 446 files: all OK
node scripts/audit-port-types.js        No port contradicts its declared type.
test/poe-inference.test.js              17 passed, 0 failed
test/poe-agreement.test.js              17 passed, 0 failed
test/catalogue-hygiene.test.js          25 passed, 0 failed
test/hardware-invariants.test.js        26 passed, 0 failed
test/import-devices-capabilities.test.js 34 passed, 0 failed
test/importer-roundtrip.test.js         26 passed, 0 failed
node scripts/regen-registry-doc.js      2,356 models across 258 vendors - no drift

Port counts and speeds are untouched: 5x (1x 2.5G + 4x 1G) and 9x 1G + 2x 10G SFP+ stand as
before, each matching the vendor Port Layout row quoted above.

Driving request: the UDW-PRO asked for in issue 130 still has no Ubiquiti source at all, so no
dream-wall Pro row is added here and issue 130 stays open.

Refs #130. `isPoePd()` matches `unifi` in `model + ' ' + label`, so every RJ45 socket on a UniFi gateway that does not declare its own `poeIn` was filled in as a 12.95 W 802.3af powered device. Two gateway rows in `devices/ubiquiti/` were still in that state; the UXG family was cleared last pass (#2121). | row | ports drawing | phantom load before | after | |---|---|---|---| | `devices/ubiquiti/UCG-Ultra.json` | wan1, port1-port4 | 64.75 W | 0 W | | `devices/ubiquiti/UDM-Pro.json` | wan1, port1-port8 | 116.55 W | 0 W | Measured through the engine, not by hand: `new Device('x', spec)` and summing `interfaces.hardware.poeIn` on `origin/main` versus this branch. ## What the vendor says **UCG-Ultra** — `https://techspecs.ui.com/unifi/cloud-gateways/ucg-ultra` (HTTP 200, 657,109 B): Port Layout `(4) 1 GbE & (1) 2.5 GbE`, Max. WAN Port Count 4, Max. Power Consumption **6.2 W**, Power Method **USB type C (5V DC/3A)**, Power Supply `USB type C (5V DC/3A) (Adapter included)`. Zero occurrences of `PoE` or `802.3` in the document. `https://store.ui.com/us/en/products/ucg-ultra` (200, 550,854 B): `PoE Support` flag **Empty**, `PoE Budget` flag **Empty**, `Max. PoE Wattage per Port by PSE` flag **Empty**. **UDM-Pro** — `https://techspecs.ui.com/unifi/cloud-gateways/udm-pro` (200, 676,527 B): Port Layout `(9) 1 GbE & (2) 10G SFP+`, Max. Power Consumption **33 W**, Power Method `(1) Universal AC input, 100-240V AC, 50/60 Hz` + `(1) USP-RPS DC input`, Power Supply `AC/DC, internal, 50W`. Zero occurrences of `PoE` or `802.3`. `https://store.ui.com/us/en/products/udm-pro` (200, 600,589 B): `PoE Support`, `PoE Budget` and `Max. PoE Wattage per Port by PSE` all flag **Empty**. So the watts were not merely unverified, they were impossible: a 6.2 W USB-C console and a 50 W AC/DC console, neither with a PSE, each "receiving" more power over copper than it can consume. ## Traps recorded * **The store page is not a single product's spec sheet.** `ucg-ultra`'s payload contains `Power Method = PoE / IEEE 802.3af` and copy for a "PoE Integrated AC Adapter" and a "PoE-powered 5G RedCap antenna". Those are cross-sell `SpecificationEntity` blocks for other SKUs; the entries that belong to this product are the ones carrying the `Empty` flags above. Reading the first `PoE` string on the page would have made this console a PD for real. * **No UCG-Ultra datasheet exists.** `https://dl.ui.com/ds/ucg-ultra.pdf` and `.../ucg-ultra_ds.pdf` both answer **302 -> `https://techspecs.ui.com`** - the same silent fallback already recorded for `uxg-fiber`, not a redirect to a PDF. * `devices/ubiquiti/UDM-Pro.json` carried **two `sourceUrl` keys**; JSON kept the last (`https://dl.ui.com/ds/udm-pro.pdf`), so the row was not uncited, but the duplicate is gone. ## Gates ``` node scripts/syntax-check.js Syntax-checked 446 files: all OK node scripts/audit-port-types.js No port contradicts its declared type. test/poe-inference.test.js 17 passed, 0 failed test/poe-agreement.test.js 17 passed, 0 failed test/catalogue-hygiene.test.js 25 passed, 0 failed test/hardware-invariants.test.js 26 passed, 0 failed test/import-devices-capabilities.test.js 34 passed, 0 failed test/importer-roundtrip.test.js 26 passed, 0 failed node scripts/regen-registry-doc.js 2,356 models across 258 vendors - no drift ``` Port counts and speeds are untouched: 5x (1x 2.5G + 4x 1G) and 9x 1G + 2x 10G SFP+ stand as before, each matching the vendor Port Layout row quoted above. Driving request: the UDW-PRO asked for in issue 130 still has no Ubiquiti source at all, so no dream-wall Pro row is added here and issue 130 stays open.
devices(ubiquiti): UCG-Ultra and UDM-Pro accept no PoE, so their copper ports are not PDs (refs #130)
Some checks failed
CI / Classify the change (pull_request) Successful in 12s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 9s
CI / Server smoke test (pull_request) Has been skipped
CI / Docker build (pull_request) Has been skipped
CI / Test (Node 20) (pull_request) Failing after 53s
CI / Test (Node 22) (pull_request) Failing after 54s
CI / Test (Node 18) (pull_request) Failing after 58s
CI / Arm auto-merge (pull_request) Successful in 3s
6534489b5b
isPoePd() matches 'unifi' in the model+label, so every silent copper port on a UniFi
gateway became a 12.95 W 802.3af draw. That put 64.75 W of phantom load on a 6.2 W
UCG-Ultra and 116.55 W on a UDM-Pro whose own supply is 50 W AC/DC.

Both rows now say poeIn: false on every RJ45 socket, cited to the vendor's own pages:
techspecs.ui.com/unifi/cloud-gateways/{ucg-ultra,udm-pro} (Power Method USB-C 5V/3A and
Universal AC + USP-RPS DC respectively, zero PoE rows in the whole document) and the
matching store rows (PoE Support / PoE Budget / Max. PoE Wattage per Port by PSE all
flagged Empty). UDM-Pro also carried a duplicate sourceUrl key; collapsed.

Refs #130.
Author
Owner

Red classification: real failing suite, not superseded, not stale. (#2124, head 6534489b, base 59a45b53)

  • Merges cleanly into current main — no CONFLICT (add/add), no files main already carries. It is not a superseded branch.
  • The vendor claim in the diff holds up on inspection (UCG-Ultra: no PoE row, USB-C 5 V power; UDM-Pro: no PoE in). The red is not the JSON being wrong — it is what the JSON removal breaks.

Failure (reproduced locally on 6534489b alone, node 26): npm test → 367 suites, 12,347 passed, 10 failed in five suites:

✗ poe.test.js              0/6   "and the PD declares what it draws (false W)"
                                   then TypeError: Cannot read properties of undefined (reading 'pd')
✗ poe-allocation-order.js  0/1   (same fixture, same crash)
✗ flow.test.js             85/91 "the PSE had power for exactly one of the two devices"
✗ show-power-inline.js     0/n   draws "Available:25.0(w) Used:0.0(w) ... the far end is not a powered device"
✗ ui-render.test.js        0/n   "and flags the port that missed out"

Root cause. UCG-Ultra is the canonical powered-device fixture in this repo, and this PR makes every one of its ports poeIn: false (correctly — isPoePd() had been inventing 12.95 W per silent port on a 6.2 W USB-C box). The fixture now has no PoE draw, so the budget/denial/flow scenarios have nothing to power:

  • test/poe.test.js:21 const PD = 'UCG-Ultra';
  • test/poe-allocation-order.test.js:29 same
  • test/flow.test.js:259 const PD = 'UCG-Ultra';
  • test/show-power-inline.test.js:46 non-plain case is UCG-Ultra
  • test/ui-render.test.js:182,212 PoE render fixtures are UCG-Ultra

Needs a decision, not a rebase. The branch is right about the hardware and the tests are right about the engine behaviour; the fixture has to move to a device the catalogue still records as PoE-fed (there are "poeIn": true entries under devices/mikrotik/, plus any capabilities: ["poe-pd"] device). Rewriting five test/ files is outside my remit (I never edit test/ as my own contribution), and picking a replacement PD changes what those suites assert, so it is a source decision.

Ask: either the catalogue loop rebases this branch and repoints the five fixtures at a genuine PD (keeping the vendor facts), or Jay says the PoE-PD inference should be keyed on something other than UCG-Ultra. Until then #2124 stays unmerged: it has a red head SHA.

**Red classification: real failing suite, not superseded, not stale.** (#2124, head 6534489b, base 59a45b53) - Merges cleanly into current main — no `CONFLICT (add/add)`, no files main already carries. It is not a superseded branch. - The vendor claim in the diff holds up on inspection (UCG-Ultra: no PoE row, USB-C 5 V power; UDM-Pro: no PoE in). The red is **not** the JSON being wrong — it is what the JSON removal breaks. **Failure (reproduced locally on 6534489b alone, node 26):** `npm test` → 367 suites, 12,347 passed, **10 failed** in five suites: ``` ✗ poe.test.js 0/6 "and the PD declares what it draws (false W)" then TypeError: Cannot read properties of undefined (reading 'pd') ✗ poe-allocation-order.js 0/1 (same fixture, same crash) ✗ flow.test.js 85/91 "the PSE had power for exactly one of the two devices" ✗ show-power-inline.js 0/n draws "Available:25.0(w) Used:0.0(w) ... the far end is not a powered device" ✗ ui-render.test.js 0/n "and flags the port that missed out" ``` **Root cause.** `UCG-Ultra` is the canonical *powered-device* fixture in this repo, and this PR makes every one of its ports `poeIn: false` (correctly — `isPoePd()` had been inventing 12.95 W per silent port on a 6.2 W USB-C box). The fixture now has no PoE draw, so the budget/denial/flow scenarios have nothing to power: - `test/poe.test.js:21` `const PD = 'UCG-Ultra';` - `test/poe-allocation-order.test.js:29` same - `test/flow.test.js:259` `const PD = 'UCG-Ultra';` - `test/show-power-inline.test.js:46` non-plain case is `UCG-Ultra` - `test/ui-render.test.js:182,212` PoE render fixtures are `UCG-Ultra` **Needs a decision, not a rebase.** The branch is right about the hardware and the tests are right about the engine behaviour; the fixture has to move to a device the catalogue still records as PoE-fed (there are `"poeIn": true` entries under `devices/mikrotik/`, plus any `capabilities: ["poe-pd"]` device). Rewriting five `test/` files is outside my remit (I never edit `test/` as my own contribution), and picking a replacement PD changes what those suites assert, so it is a source decision. **Ask:** either the catalogue loop rebases this branch and repoints the five fixtures at a genuine PD (keeping the vendor facts), or Jay says the PoE-PD inference should be keyed on something other than `UCG-Ultra`. Until then #2124 stays unmerged: it has a red head SHA.
Author
Owner

Surgery re-check after four merges (2026-11-19 09:4xZ) — still red, and the decision is now one-sided.

Head unchanged at 6534489b, now 7 behind main. Merged it onto current main locally (clean, no conflicts) and re-ran the full suite: 368 suites, 12,424 assertions, 10 failed — byte-for-byte the same ten failures in poe, poe-allocation-order, flow, show-power-inline, ui-render as at 09:16Z. The three PRs that landed meanwhile (#2128, #2126, #2127) do not touch them.

docs/plans/poe-pd-inference.md landed as 9ac4fccb (#2129) since my last comment, and it answers the question I escalated: it measures UCG-Ultra at 0 PSE / 5 PD → 64.75 W from engine/media.js:1104-1107 promoting a device-level isPoePd() claim to every copper port, with no device-shape gate and no poeInDeclared counterpart to poeOutDeclared. So this PR's JSON is right and the fixtures are right; the engine inference is what must change, and the plan already names the files (media.js:1086-1116, device.js:1433/:1651-1658, simulator.js:2555-2612, web/faceplate.js:630, web/app.js:2202, docs/DEVICE_SCHEMA.md:204, test/poe.test.js:21).

Not merging it, per the red-PR rule — not superseded (merges cleanly, no add/add), not staleness. Two options and both are source changes, so neither is mine:

  1. land the plan's engine change, then this PR is green as-is; or
  2. repoint the fixture (test/poe.test.js:21 and four siblings) to a genuine PD and let the inferred ports stand.

Needs Jay's decision: (1) or (2). Option 1 is the plan's own conclusion and fixes 448 devices' worth of inferred ports; option 2 unblocks this PR in one commit and leaves the guess in place.

**Surgery re-check after four merges (2026-11-19 09:4xZ) — still red, and the decision is now one-sided.** Head unchanged at `6534489b`, now 7 behind main. Merged it onto current main locally (clean, no conflicts) and re-ran the full suite: **368 suites, 12,424 assertions, 10 failed** — byte-for-byte the same ten failures in `poe`, `poe-allocation-order`, `flow`, `show-power-inline`, `ui-render` as at 09:16Z. The three PRs that landed meanwhile (#2128, #2126, #2127) do not touch them. `docs/plans/poe-pd-inference.md` landed as `9ac4fccb` (#2129) since my last comment, and it answers the question I escalated: it measures `UCG-Ultra` at **0 PSE / 5 PD → 64.75 W** from `engine/media.js:1104-1107` promoting a device-level `isPoePd()` claim to every copper port, with no device-shape gate and no `poeInDeclared` counterpart to `poeOutDeclared`. So this PR's JSON is right and the fixtures are right; the engine inference is what must change, and the plan already names the files (`media.js:1086-1116`, `device.js:1433`/`:1651-1658`, `simulator.js:2555-2612`, `web/faceplate.js:630`, `web/app.js:2202`, `docs/DEVICE_SCHEMA.md:204`, `test/poe.test.js:21`). **Not merging it**, per the red-PR rule — not superseded (merges cleanly, no add/add), not staleness. Two options and both are source changes, so neither is mine: 1. land the plan's engine change, then this PR is green as-is; or 2. repoint the fixture (`test/poe.test.js:21` and four siblings) to a genuine PD and let the inferred ports stand. **Needs Jay's decision: (1) or (2).** Option 1 is the plan's own conclusion and fixes 448 devices' worth of inferred ports; option 2 unblocks this PR in one commit and leaves the guess in place.
Author
Owner

The catalogue change was right; five suites had been using UCG-Ultra as the powered far end (poe, poe-allocation-order, flow, show-power-inline, ui-render), so removing the phantom 12.95 W left them with no PD at all. Swapped that fixture for a genuine 802.3af PD (MikroTik SXTsq 5 ac, 12.95 W, single copper eth1, L3) - same arithmetic, same assertions. poe.test.js now also asserts the point of the PR: both gateways declare poeIn: false, neither port enters the allocation table, and the PSE spends 0 W on them.

Full suite: 367 suites, 12,470 assertions, 0 failed. node scripts/syntax-check.js: 446 files OK.

The catalogue change was right; five suites had been using `UCG-Ultra` as the powered far end (`poe`, `poe-allocation-order`, `flow`, `show-power-inline`, `ui-render`), so removing the phantom 12.95 W left them with no PD at all. Swapped that fixture for a genuine 802.3af PD (MikroTik SXTsq 5 ac, 12.95 W, single copper `eth1`, L3) - same arithmetic, same assertions. `poe.test.js` now also asserts the point of the PR: both gateways declare `poeIn: false`, neither port enters the allocation table, and the PSE spends 0 W on them. Full suite: 367 suites, 12,470 assertions, 0 failed. `node scripts/syntax-check.js`: 446 files OK.
test(poe): the fixtures power a device that actually draws PoE
All checks were successful
CI / Classify the change (pull_request) Successful in 8s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 9s
CI / Test (Node 20) (pull_request) Successful in 51s
CI / Test (Node 18) (pull_request) Successful in 55s
CI / Server smoke test (pull_request) Successful in 7s
CI / Test (Node 22) (pull_request) Successful in 51s
CI / Docker build (pull_request) Successful in 17s
CI / Arm auto-merge (pull_request) Successful in 3s
e57785d92e
#2124 made UCG-Ultra and UDM-Pro say `poeIn: false`, which is what their
datasheets say - but five suites had been using UCG-Ultra as the powered far
end: poe, poe-allocation-order, flow, show-power-inline, ui-render. With the
phantom 12.95 W gone they had no PD at all, and three of them died on
`sim.poe[0].ports` being undefined.

The fixture was the thing that was wrong, not the catalogue. A MikroTik SXTsq
5 ac (RBSXTsqG-5acD) is a genuine 802.3af PD at 12.95 W with one copper eth1
and an L3 stack, so it can still take an IP and ping. Same arithmetic, same
assertions, one device swap plus the port name.

poe.test.js also keeps the assertion this issue was about: both gateways
declare `poeIn: false`, the allocation table holds neither port and the PSE
spends 0 W on them.
forgejo-actions scheduled this pull request to auto merge when all checks succeed 2026-11-19 10:15:22 +00:00
jaylouisw force-pushed catalogue/1119-0852 from e57785d92e
All checks were successful
CI / Classify the change (pull_request) Successful in 8s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 9s
CI / Test (Node 20) (pull_request) Successful in 51s
CI / Test (Node 18) (pull_request) Successful in 55s
CI / Server smoke test (pull_request) Successful in 7s
CI / Test (Node 22) (pull_request) Successful in 51s
CI / Docker build (pull_request) Successful in 17s
CI / Arm auto-merge (pull_request) Successful in 3s
to 29668edbec
All checks were successful
CI / Classify the change (pull_request) Successful in 12s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 7s
CI / Server smoke test (pull_request) Successful in 10s
CI / Docker build (pull_request) Successful in 22s
CI / Test (Node 20) (pull_request) Successful in 1m0s
CI / Test (Node 22) (pull_request) Successful in 1m0s
CI / Test (Node 18) (pull_request) Successful in 1m6s
CI / Arm auto-merge (pull_request) Successful in 3s
2026-11-19 10:49:45 +00:00
Compare
jaylouisw deleted branch catalogue/1119-0852 2026-11-19 10:51:40 +00:00
Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
jaylouisw/packetforge!2124
No description provided.