devices(ubiquiti): three Enterprise Campus PoE rows served 0 W from a 2,150 W chassis #2249

Merged
jaylouisw merged 1 commit from catalogue/1120-0252 into main 2026-11-20 03:45:23 +00:00
Owner

Does not close #130 — the UDW-PRO blocker is unchanged (see the comment on that issue; no Ubiquiti-hosted artefact names the SKU). This PR repairs the Enterprise Campus PoE family instead, which is what the last pass on #130 recorded as the next job.

The defect

ecs-48-poe.json and ecs-24s-poe.json declared PoE+++ on 48 and 24 ports and carried no poe.budgetW. The loader therefore named the budget from the model string, found no wattage, and served 0 W while the description beside each claimed 2,150 W and 1,050 W. Same defect class as #2235, #2175 and #2244.

before                                   after
ecs-48-poe   budget=0W    src=name       budget=2150W src=research  jack=52 pse=48
ecs-24s-poe  budget=0W    src=name       budget=1050W src=research  jack=32 pse=24
ecs-24-poe   budget=1050W src=research   unchanged                    jack=26 pse=24
ecs-48s-poe  budget=2150W src=research   unchanged                    jack=56 pse=48

Per model

ECS-48-PoE (devices/ubiquiti/ecs-48-poe.json) — Vendor prints availability as four rows against (2) 1,200 W modules: 200-240 V shared 2,150 W / 100-120 V shared 1,750 W / 200-240 V redundant 950 W / 100-120 V redundant 750 W. 2,150 W modelled, all four conditions in poe.note. Port layout is three rows (16 x 2.5G, 32 x 10G, 4 x SFP28); no Console and no OOB port on this row, so 52 jacks is the whole box. Both RJ45 banks were missing the 10M rung.

ECS-24S-PoE (devices/ubiquiti/ecs-24s-poe.json) — Six availability rows across two PSU options. budgetW is 1,050 W, the as-shipped (2) 600 W shared figure, matching the convention already in ecs-24-poe.json; the page marks the 1,200 W modules "sold separately", so 2,250 W is recorded in poe.note as a condition a lab must state, not as the ceiling. Plus: the serial console was absent (31 of 32 jacks - nothing could be cabled to the out-of-band path), the 1G OOB port and both RJ45 banks were missing 10M, and the QSFP28 pair now carries the vendor note "Stacking port only, does not support uplink. Breakout cables are not supported".

ECS-24-PoE (devices/ubiquiti/ecs-24-poe.json) — budget was already correct; 10M rung added to both RJ45 banks, notes added.

ECS-48S-PoE (devices/ubiquiti/ecs-48s-poe.json) — finishing #2244: the 16 x 10G bank was the one bank left without the 10M rung the page prints.

All four / phantom PSE. The engine credits every RJ45 jack on a poe-pse switch with 15.4 W when the spec is silent, which put a PSE on each row's OOB management port - a port the vendor annotates "OOB Management port only". eth0 on both S rows now declares poeOut: false, which is what makes pse above equal the vendor's own "PoE Ports: 48 / 24" row instead of 49 / 25.

Evidence

Every field traces to the vendor's server-rendered spec table, byte counts recorded in each spec's notes:

URL HTTP bytes
https://techspecs.ui.com/unifi/switching/ecs-48-poe 200 663,600
https://techspecs.ui.com/unifi/switching/ecs-24s-poe 200 677,049
https://techspecs.ui.com/unifi/switching/ecs-24-poe 200 663,913
https://techspecs.ui.com/unifi/switching/ecs-48s-poe 200 676,185

Port-layout rows are quoted verbatim in the notes; the tooltip annotations ("OOB Management port only", "Stacking port only...", "1,200W power modules sold separately", "802.3bt Type 4, Class 8") are quoted from the page's own spec JSON.

Gates

node scripts/syntax-check.js          -> 475 files: all OK
node scripts/audit-port-types.js      -> No port contradicts its declared type.
node test/hardware-invariants.test.js -> 26 passed, 0 failed
importer --all (dry run)              -> 4 specs scanned, 0 rejected, capability tripwire silent
npm test                              -> 395 suites, 13,212 assertions, 0 failed

docs/CONNECTORS.md regenerated (2,355 models, 29,726 sockets, 973 distinct port shapes); docs/DEVICE_REGISTRY.md unchanged.

Does not close #130 — the UDW-PRO blocker is unchanged (see the comment on that issue; no Ubiquiti-hosted artefact names the SKU). This PR repairs the Enterprise Campus PoE family instead, which is what the last pass on #130 recorded as the next job. ## The defect `ecs-48-poe.json` and `ecs-24s-poe.json` declared PoE+++ on 48 and 24 ports and carried no `poe.budgetW`. The loader therefore named the budget from the model string, found no wattage, and served **0 W** while the `description` beside each claimed 2,150 W and 1,050 W. Same defect class as #2235, #2175 and #2244. ``` before after ecs-48-poe budget=0W src=name budget=2150W src=research jack=52 pse=48 ecs-24s-poe budget=0W src=name budget=1050W src=research jack=32 pse=24 ecs-24-poe budget=1050W src=research unchanged jack=26 pse=24 ecs-48s-poe budget=2150W src=research unchanged jack=56 pse=48 ``` ## Per model **ECS-48-PoE** (`devices/ubiquiti/ecs-48-poe.json`) — Vendor prints availability as four rows against (2) 1,200 W modules: 200-240 V shared 2,150 W / 100-120 V shared 1,750 W / 200-240 V redundant 950 W / 100-120 V redundant 750 W. 2,150 W modelled, all four conditions in `poe.note`. Port layout is three rows (16 x 2.5G, 32 x 10G, 4 x SFP28); no Console and no OOB port on this row, so 52 jacks is the whole box. Both RJ45 banks were missing the 10M rung. **ECS-24S-PoE** (`devices/ubiquiti/ecs-24s-poe.json`) — Six availability rows across two PSU options. `budgetW` is **1,050 W**, the as-shipped (2) 600 W shared figure, matching the convention already in `ecs-24-poe.json`; the page marks the 1,200 W modules "sold separately", so 2,250 W is recorded in `poe.note` as a condition a lab must state, not as the ceiling. Plus: the **serial console was absent** (31 of 32 jacks - nothing could be cabled to the out-of-band path), the 1G OOB port and both RJ45 banks were missing 10M, and the QSFP28 pair now carries the vendor note "Stacking port only, does not support uplink. Breakout cables are not supported". **ECS-24-PoE** (`devices/ubiquiti/ecs-24-poe.json`) — budget was already correct; 10M rung added to both RJ45 banks, notes added. **ECS-48S-PoE** (`devices/ubiquiti/ecs-48s-poe.json`) — finishing #2244: the 16 x 10G bank was the one bank left without the 10M rung the page prints. **All four / phantom PSE.** The engine credits every RJ45 jack on a `poe-pse` switch with 15.4 W when the spec is silent, which put a PSE on each row's OOB management port - a port the vendor annotates "OOB Management port only". `eth0` on both S rows now declares `poeOut: false`, which is what makes `pse` above equal the vendor's own "PoE Ports: 48 / 24" row instead of 49 / 25. ## Evidence Every field traces to the vendor's server-rendered spec table, byte counts recorded in each spec's `notes`: | URL | HTTP | bytes | |---|---|---| | https://techspecs.ui.com/unifi/switching/ecs-48-poe | 200 | 663,600 | | https://techspecs.ui.com/unifi/switching/ecs-24s-poe | 200 | 677,049 | | https://techspecs.ui.com/unifi/switching/ecs-24-poe | 200 | 663,913 | | https://techspecs.ui.com/unifi/switching/ecs-48s-poe | 200 | 676,185 | Port-layout rows are quoted verbatim in the notes; the tooltip annotations ("OOB Management port only", "Stacking port only...", "1,200W power modules sold separately", "802.3bt Type 4, Class 8") are quoted from the page's own spec JSON. ## Gates ``` node scripts/syntax-check.js -> 475 files: all OK node scripts/audit-port-types.js -> No port contradicts its declared type. node test/hardware-invariants.test.js -> 26 passed, 0 failed importer --all (dry run) -> 4 specs scanned, 0 rejected, capability tripwire silent npm test -> 395 suites, 13,212 assertions, 0 failed ``` `docs/CONNECTORS.md` regenerated (2,355 models, 29,726 sockets, 973 distinct port shapes); `docs/DEVICE_REGISTRY.md` unchanged.
devices(ubiquiti): three Enterprise Campus PoE rows served 0 W from a 2,150 W chassis
All checks were successful
CI / Classify the change (pull_request) Successful in 9s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 26s
CI / Test (Node 20) (pull_request) Successful in 1m5s
CI / Test (Node 22) (pull_request) Successful in 1m4s
CI / Test (Node 18) (pull_request) Successful in 1m11s
CI / Server smoke and Docker build (pull_request) Successful in 18s
CI / Arm auto-merge (pull_request) Successful in 4s
f4030557f8
The ECS-48-PoE and ECS-24S-PoE specs declared PoE+++ on 48 and 24 ports and no
poe.budgetW, so the loader named the budget from the model string, found no
wattage and served 0 W while the description beside each claimed 2,150 W and
1,050 W. Same defect class as #2235, #2175 and #2244.

* ECS-48-PoE   - budget 2,150 W, four availability rows in poe.note
* ECS-24S-PoE  - budget 1,050 W (as-shipped 2x 600 W PSUs), six rows in
                 poe.note, 2,250 W only with the separately sold 1,200 W
                 modules
* ECS-24-PoE   - budget was already right; 10M rung added to both banks

Also from the same three pages: the 24S serial console was missing (31 of 32
jacks), the 10M rung was absent on every RJ45 bank in all four files, and the
OOB management port on the two S rows now declares poeOut: false - the engine
credits every RJ45 jack on a poe-pse switch with 15.4 W, which put a phantom
PSE on a port the vendor marks 'OOB Management port only'.

Sources: techspecs.ui.com/unifi/switching/{ecs-48-poe,ecs-24s-poe,ecs-24-poe,
ecs-48s-poe}, read 2026-11-20, byte counts in each spec's notes.
forgejo-actions scheduled this pull request to auto merge when all checks succeed 2026-11-20 03:14:52 +00:00
jaylouisw force-pushed catalogue/1120-0252 from f4030557f8
All checks were successful
CI / Classify the change (pull_request) Successful in 9s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 26s
CI / Test (Node 20) (pull_request) Successful in 1m5s
CI / Test (Node 22) (pull_request) Successful in 1m4s
CI / Test (Node 18) (pull_request) Successful in 1m11s
CI / Server smoke and Docker build (pull_request) Successful in 18s
CI / Arm auto-merge (pull_request) Successful in 4s
to bbe25d20f1
All checks were successful
CI / Classify the change (pull_request) Successful in 13s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 13s
CI / Server smoke and Docker build (pull_request) Has been skipped
CI / Test (Node 20) (pull_request) Successful in 1m1s
CI / Test (Node 18) (pull_request) Successful in 1m10s
CI / Test (Node 22) (pull_request) Successful in 1m1s
CI / Arm auto-merge (pull_request) Successful in 5s
2026-11-20 03:40:22 +00:00
Compare
jaylouisw deleted branch catalogue/1120-0252 2026-11-20 03:45:25 +00:00
Author
Owner

Rebased onto main (f86c9a27) and squash-merged as 84426ad8. The only conflict was the catalogue totals sentence in docs/CONNECTORS.md, which #2251 had already moved: rather than pick a side, node scripts/regen-registry-doc.js rewrote it from the merged tree -> 2,355 models, 29,729 sockets, 972 distinct port shapes. Re-ran audit-port-types, check-control-bytes, doc-catalog-counts (43), faceplate-catalogue (14), hardware-invariants (26), poe (21), validation (22) - all green before the force-push; head bbe25d20 went 7/7.

Rebased onto main (f86c9a27) and squash-merged as 84426ad8. The only conflict was the catalogue totals sentence in `docs/CONNECTORS.md`, which #2251 had already moved: rather than pick a side, `node scripts/regen-registry-doc.js` rewrote it from the merged tree -> **2,355 models, 29,729 sockets, 972 distinct port shapes**. Re-ran audit-port-types, check-control-bytes, doc-catalog-counts (43), faceplate-catalogue (14), hardware-invariants (26), poe (21), validation (22) - all green before the force-push; head bbe25d20 went 7/7.
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!2249
No description provided.