devices(mikrotik): nine PoE switches, seven wrong port ranges and five published budgets #2294

Merged
jaylouisw merged 1 commit from catalogue/1120-1752 into main 2026-11-20 18:07:54 +00:00
Owner

Refs #803, #130 (both still blocked; see the notes at the end).

MikroTik's product pages publish the PoE census port by port (PoE-out ports, Number of 1G Ethernet ports with PoE-out, Total output power), and the catalogue disagreed with them. The engine infers "every RJ45 port of a poe-pse switch sources 15.4 W" (isPoePse(), engine/media.js), which is exactly wrong on a reverse-PoE box: it is powered from its subscribers.

Wrong ports marked as sources

spec vendor row was now
css610-1gi-7r-2s-out (netPower Lite 7R) PoE-out ports Ether8, 7 x Reverse PoE in 8 PSE 1 PSE
crs318-1fi-15fr-2s-out (netPower 15FR) PoE-out ports Ether15, 15 reverse PoE 16 PSE 1 PSE
rb260gsp PoE-out ports Ether2-Ether5 (Ether1 = 11-30 V in) 5 PSE 4 PSE
gperx6 PoE-out ports Ether2-Ether3, Ether1 = 802.3bt in eth1-2 PoE+ eth2-3 passive
crs320-8p-8b-4s-rm Ether1-Ether8 (af/at), Ether9-Ether16 (bt), 17 GbE 17 PSE 16 PSE (eth17 mgmt)
crs418-8p-8g-2s-rm PoE-out ports Ether9-Ether16 17 PSE 8 PSE
crs354-48p-4s-2q-rm PoE-out ports Ether1-Ether48 + separate 10/100 Ethernet ports 1 49 PSE 48 PSE

Missing copper ports

crs328-24p-4s-rm builds its 24 copper ports from a portTemplate that carried no poeOut, so a 24-port PoE switch modelled as unpowered. The template now sources and the spec gains poe-pse / 802.3af / 802.3at.

Published budgets (Total output power)

  • css610-8p-2s-in — 140 W
  • crs320-8p-8b-4s-rm — 963 W (17.82 A)
  • crs418-8p-8g-2s-rm — 150 W (5.4 A)
  • crs354-48p-4s-2q-rm — 700 W (27 A @26 V / 13.2 A @53 V)
  • gperx6 — 150 W, pass-through, and 802.3at in the capability list corrected to 802.3bt (the input standard; the output is fixed-voltage passive, so the two PoE+ classes became passive)

Arithmetic not done

NetPower Lite 7R, netPower 15FR, RB260GSP, CRS328 and CRS112-8P-4S-IN state current limits only (Max total out in A against a voltage window) and never watts. Multiplying 0.6 A by 57 V is arithmetic MikroTik did not do, so those four carry poe.budgetUnknown: true with the quoted limit in poe.note instead of a made-up figure.

Evidence

All pages read 2026-11-20, HTTP 200, Mozilla/5.0 … Chrome/126 UA:

/product/<slug> is case-sensitive, uses underscores, and 404s (13,238 B) on the hyphenised guesses for every one of these; https://mikrotik.com/sitemap.xml (200, 48,450 B, 653 product URLs) is the only reliable way to find the slug. research/mikrotik-poe-out-census.md records the rows verbatim.

Gates

  • node scripts/syntax-check.js — 490 files, all OK
  • node scripts/audit-port-types.js — no port contradicts its declared type
  • node scripts/audit-poe-budget.js — silent 69 → 61, unknown-on-purpose 17 → 21, orphan 0
  • node test/hardware-invariants.test.js 26/26, spec-vs-description 3/3, poe-inference 19/19, catalogue-hygiene 25/25, faceplate 14/14
  • node scripts/run-tests.js — 410 suites, 13,604 assertions, 0 failed
  • docs/DEVICE_REGISTRY.md regenerated (no count or shape change: 2,355 models / 258 vendors / 29,735 sockets / 974 shapes)

Queue status (unchanged)

#803 (OptiXtrans E6616X) stays blocked on any service-board part number valid in the X subrack, with its front-panel census. #130 (UDW-PRO) stays blocked: no Ubiquiti-published artefact names UDW-PRO. Neither was re-probed this run; this PR does not close either.

Refs #803, #130 (both still blocked; see the notes at the end). MikroTik's product pages publish the PoE census port by port (`PoE-out ports`, `Number of 1G Ethernet ports with PoE-out`, `Total output power`), and the catalogue disagreed with them. The engine infers "every RJ45 port of a `poe-pse` switch sources 15.4 W" (`isPoePse()`, `engine/media.js`), which is exactly wrong on a reverse-PoE box: it is powered *from* its subscribers. ## Wrong ports marked as sources | spec | vendor row | was | now | |---|---|---|---| | `css610-1gi-7r-2s-out` (netPower Lite 7R) | `PoE-out ports Ether8`, 7 x Reverse PoE in | 8 PSE | 1 PSE | | `crs318-1fi-15fr-2s-out` (netPower 15FR) | `PoE-out ports Ether15`, 15 reverse PoE | 16 PSE | 1 PSE | | `rb260gsp` | `PoE-out ports Ether2-Ether5` (Ether1 = 11-30 V in) | 5 PSE | 4 PSE | | `gperx6` | `PoE-out ports Ether2-Ether3`, Ether1 = 802.3bt in | eth1-2 PoE+ | eth2-3 passive | | `crs320-8p-8b-4s-rm` | `Ether1-Ether8 (af/at), Ether9-Ether16 (bt)`, 17 GbE | 17 PSE | 16 PSE (eth17 mgmt) | | `crs418-8p-8g-2s-rm` | `PoE-out ports Ether9-Ether16` | 17 PSE | 8 PSE | | `crs354-48p-4s-2q-rm` | `PoE-out ports Ether1-Ether48` + separate `10/100 Ethernet ports 1` | 49 PSE | 48 PSE | ## Missing copper ports `crs328-24p-4s-rm` builds its 24 copper ports from a `portTemplate` that carried no `poeOut`, so a 24-port PoE switch modelled as unpowered. The template now sources and the spec gains `poe-pse` / `802.3af` / `802.3at`. ## Published budgets (`Total output power`) * `css610-8p-2s-in` — 140 W * `crs320-8p-8b-4s-rm` — 963 W (17.82 A) * `crs418-8p-8g-2s-rm` — 150 W (5.4 A) * `crs354-48p-4s-2q-rm` — 700 W (27 A @26 V / 13.2 A @53 V) * `gperx6` — 150 W, pass-through, and `802.3at` in the capability list corrected to `802.3bt` (the input standard; the output is fixed-voltage passive, so the two `PoE+` classes became `passive`) ## Arithmetic not done NetPower Lite 7R, netPower 15FR, RB260GSP, CRS328 and CRS112-8P-4S-IN state **current** limits only (`Max total out` in A against a voltage window) and never watts. Multiplying 0.6 A by 57 V is arithmetic MikroTik did not do, so those four carry `poe.budgetUnknown: true` with the quoted limit in `poe.note` instead of a made-up figure. ## Evidence All pages read 2026-11-20, HTTP 200, `Mozilla/5.0 … Chrome/126` UA: * https://mikrotik.com/product/css610_8p_2s_in (351,350 B) * https://mikrotik.com/product/crs320_8p_8b_4s_rm (364,538 B) * https://mikrotik.com/product/crs418_8p_8g_2s_rm (419,515 B) * https://mikrotik.com/product/crs354_48p_4s_2q_rm (394,160 B) * https://mikrotik.com/product/gperx6 (353,837 B) * https://mikrotik.com/product/netpower_lite_7r (356,904 B) * https://mikrotik.com/product/crs318_1fi_15fr_2s_out (380,113 B) * https://mikrotik.com/product/RB260GSP (344,158 B) * https://mikrotik.com/product/crs328_24p_4s_rm (368,617 B) `/product/<slug>` is case-sensitive, uses underscores, and 404s (13,238 B) on the hyphenised guesses for every one of these; `https://mikrotik.com/sitemap.xml` (200, 48,450 B, 653 product URLs) is the only reliable way to find the slug. `research/mikrotik-poe-out-census.md` records the rows verbatim. ## Gates * `node scripts/syntax-check.js` — 490 files, all OK * `node scripts/audit-port-types.js` — no port contradicts its declared type * `node scripts/audit-poe-budget.js` — silent 69 → 61, unknown-on-purpose 17 → 21, orphan 0 * `node test/hardware-invariants.test.js` 26/26, `spec-vs-description` 3/3, `poe-inference` 19/19, `catalogue-hygiene` 25/25, `faceplate` 14/14 * `node scripts/run-tests.js` — 410 suites, 13,604 assertions, 0 failed * `docs/DEVICE_REGISTRY.md` regenerated (no count or shape change: 2,355 models / 258 vendors / 29,735 sockets / 974 shapes) ## Queue status (unchanged) #803 (OptiXtrans E6616X) stays blocked on any service-board part number valid in the X subrack, with its front-panel census. #130 (UDW-PRO) stays blocked: no Ubiquiti-published artefact names `UDW-PRO`. Neither was re-probed this run; this PR does not close either.
devices(mikrotik): nine PoE switches, seven wrong port ranges and five published budgets
All checks were successful
CI / Classify the change (pull_request) Successful in 14s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 12s
CI / Server smoke and Docker build (pull_request) Successful in 25s
CI / Test (Node 22) (pull_request) Successful in 59s
CI / Test (Node 20) (pull_request) Successful in 1m1s
CI / Test (Node 18) (pull_request) Successful in 1m5s
CI / Arm auto-merge (pull_request) Successful in 5s
7004ef48f1
MikroTik's product pages publish the PoE census port by port, and the catalogue
disagreed with them. The engine infers "every RJ45 port of a `poe-pse` switch
sources 15.4 W" (isPoePse()), which is exactly wrong on a reverse-PoE box.

Wrong port ranges, from the vendor's `PoE-out ports` row:
  netPower Lite 7R     Ether8 only      (was 8 PSE; 7 ports are PoE-*in*)
  netPower 15FR        Ether15 only     (was 16 PSE; 15 ports are PoE-*in*)
  RB260GSP             Ether2-Ether5    (was 5; Ether1 is the 11-30 V input)
  GPERx6               Ether2-Ether3    (was Ether1-Ether2; Ether1 is 802.3bt in)
  CRS320-8P-8B-4S+RM   Ether1-Ether16   (eth17 is the management port)
  CRS418-8P-8G-2S+RM   Ether9-Ether16   (the 8P names the PoE ports, not the first 8)
  CRS354-48P-4S+2Q+RM  Ether1-Ether48   (the 100M management port is not a PSE)

Published watts, `Total output power`: CSS610-8P-2S+IN 140, CRS320 963,
CRS418 150, CRS354 700, GPERx6 150.

CRS328-24P-4S+RM: its 24 copper ports come from a port template that carried no
`poeOut`, so a 24-port PoE switch modelled as unpowered. Template now sources,
capabilities gain poe-pse/802.3af/802.3at.

NetPower Lite 7R, netPower 15FR, RB260GSP, CRS328 and CRS112 publish current
limits only (A against a voltage window), never watts - they are marked
`poe.budgetUnknown` with the quoted limit in the note rather than multiplied out.

Sources: mikrotik.com/product/{css610_8p_2s_in,crs320_8p_8b_4s_rm,
crs418_8p_8g_2s_rm,crs354_48p_4s_2q_rm,gperx6,netpower_lite_7r,
crs318_1fi_15fr_2s_out,rb260gsp,crs328_24p_4s_rm}, read 2026-11-20.
research/mikrotik-poe-out-census.md records the rows and the slug trap
(/product/<slug> is case-sensitive, underscores, and 404s on hyphen guesses).

Gates: syntax-check 490 files OK; audit-port-types no contradiction;
audit-poe-budget silent 69 -> 61, unknown 17 -> 21; run-tests 410 suites,
13,604 assertions, 0 failed.
forgejo-actions scheduled this pull request to auto merge when all checks succeed 2026-11-20 18:04:03 +00:00
jaylouisw deleted branch catalogue/1120-1752 2026-11-20 18:07:57 +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!2294
No description provided.