catalogue(planet): six GS-4210 L2/L4 managed gigabit switches #1603

Closed
jaylouisw wants to merge 1 commit from catalogue/1116-0425 into main
Owner

Adds the PLANET GS-4210 L2/L4 managed gigabit access family (vendor planet, previously 5 devices, all media converters and PoE extenders). Six specs, one hardware revision each, every figure traceable to a PLANET datasheet or product page.

model ports PoE fabric source
GS-4210-8P2S 8x1G PoE+ 30 W + 2x 100/1000X SFP + RJ45 console 140 W @25 C (120 W @50 C) 20 Gbps, 14.88 Mpps datasheet v3
GS-4210-16P2S 16x1G PoE+ 30 W + 2x 100/1000X SFP + console 220 W 36 Gbps, 26.78 Mpps product page hw v2
GS-4210-24P2S 24x1G PoE+ 30 W + 2x 100/1000/2500X SFP + console 370 W 58 Gbps, 43.1 Mpps product page hw v4
GS-4210-24T4S 24x1G + 4x 1000/2500X SFP + console, no PoE n/a 68 Gbps, 50.5 Mpps product page hw v2
GS-4210-48T4S 48x1G + 4x 100/1000X SFP (49-52) + console, no PoE n/a 104 Gbps, 77.38 Mpps product page hw v3
GS-4210-48P4S 48x1G PoE+ 36 W + 4x 100/1000X SFP + console 600 W @25 C (500 W @50 C) 104 Gbps datasheet v3

Discrepancies stated rather than smoothed over

PLANET restates the same model number with different electrical figures across hardware revisions, so each entry names the revision it models in notes:

  • GS-4210-24P2S: v2 datasheet = 52 Gbps fabric and 2x 100/1000X SFP; series quick-install guide = 300 W budget; current v4 page = 58 Gbps, 2.5G SFP slots, 370 W. This PR models v4 and does not claim 2.5G on the v2 hardware.
  • GS-4210-16P2S: a 2025-10-31 v3 sheet lists 42 Gbps, 240 W, 16K MAC and 1G/2.5G SFPs; the v2 page modelled here lists 36 Gbps, 220 W, 8K MAC and 1G-only SFPs.
  • GS-4210-8P2S: earlier sheet = 120 W budget and 36 W/port; v3 sheet = 140 W and 30 W/port.
  • GS-4210-24T4S: the 2021 user manual states 56 Gbps for hardware v1; the v2 page states 68 Gbps. Modelled at v2.
  • GS-4210-24P4C / 24PL4C deliberately not imported this run: 220 W/440 W on the 2019 combo-port sheet vs 370 W/650 W on the current page, and the combo ports need the shared-cage modelling decision that these six did not raise.

Extend mode (10 Mbps per port over 250 m) is documented in notes for the 48P4S but not modelled - the port is still declared at 1G.

Validation

  • node scripts/import-devices.js --dry-run over the batch: Scanned 1 file(s), 6 spec(s) / Would add: 6 duplicates: 0 rejected: 0
  • node scripts/import-devices.js --apply: added 6, no capability tokens dropped (all are engine tokens: l2-forward, vlan, lacp, stp, cli, http-server, poe-pse)
  • node scripts/audit-port-types.js: No port contradicts its declared type. (exit 0)
  • node scripts/audit-devices.js: No hard defects.
  • node scripts/syntax-check.js: 333 files, all OK
  • node --test test/importer-roundtrip.test.js test/import-devices-capabilities.test.js test/fuzz-importer.test.js: 26 passed, 0 failed
  • node --test test/catalogue-hygiene.test.js: 25 passed, 0 failed
  • Engine load check: 48P4S resolves 53 interfaces, 48 with 36 W poeOut, port 49 cages as SFP/optical.
Adds the PLANET **GS-4210** L2/L4 managed gigabit access family (vendor `planet`, previously 5 devices, all media converters and PoE extenders). Six specs, one hardware revision each, every figure traceable to a PLANET datasheet or product page. | model | ports | PoE | fabric | source | |---|---|---|---|---| | GS-4210-8P2S | 8x1G PoE+ 30 W + 2x 100/1000X SFP + RJ45 console | 140 W @25 C (120 W @50 C) | 20 Gbps, 14.88 Mpps | [datasheet v3](https://www.planet.com.tw/storage/products/49452/C-GS-4210-8P2Sv3_L.pdf) | | GS-4210-16P2S | 16x1G PoE+ 30 W + 2x 100/1000X SFP + console | 220 W | 36 Gbps, 26.78 Mpps | [product page hw v2](https://www.planet.com.tw/en/product/gs-4210-16p2s-v2) | | GS-4210-24P2S | 24x1G PoE+ 30 W + 2x 100/1000/2500X SFP + console | 370 W | 58 Gbps, 43.1 Mpps | [product page hw v4](https://www.planet.com.tw/en/product/gs-4210-24p2s-v4) | | GS-4210-24T4S | 24x1G + 4x 1000/2500X SFP + console, no PoE | n/a | 68 Gbps, 50.5 Mpps | [product page hw v2](https://www.planet.com.tw/en/product/gs-4210-24t4s-gs-4210-24t4sr-v2) | | GS-4210-48T4S | 48x1G + 4x 100/1000X SFP (49-52) + console, no PoE | n/a | 104 Gbps, 77.38 Mpps | [product page hw v3](https://www.planet.com.tw/en/product/gs-4210-48t4s-v3) | | GS-4210-48P4S | 48x1G PoE+ 36 W + 4x 100/1000X SFP + console | 600 W @25 C (500 W @50 C) | 104 Gbps | [datasheet v3](https://www.planet.com.tw/en/product/gs-4210-48p4s-v3) | ## Discrepancies stated rather than smoothed over PLANET restates the same model number with different electrical figures across hardware revisions, so each entry names the revision it models in `notes`: - **GS-4210-24P2S**: v2 datasheet = 52 Gbps fabric and 2x 100/1000X SFP; series quick-install guide = 300 W budget; current v4 page = 58 Gbps, 2.5G SFP slots, 370 W. This PR models v4 and does not claim 2.5G on the v2 hardware. - **GS-4210-16P2S**: a 2025-10-31 v3 sheet lists 42 Gbps, 240 W, 16K MAC and 1G/2.5G SFPs; the v2 page modelled here lists 36 Gbps, 220 W, 8K MAC and 1G-only SFPs. - **GS-4210-8P2S**: earlier sheet = 120 W budget and 36 W/port; v3 sheet = 140 W and 30 W/port. - **GS-4210-24T4S**: the 2021 user manual states 56 Gbps for hardware v1; the v2 page states 68 Gbps. Modelled at v2. - **GS-4210-24P4C / 24PL4C** deliberately not imported this run: 220 W/440 W on the 2019 combo-port sheet vs 370 W/650 W on the current page, and the combo ports need the shared-cage modelling decision that these six did not raise. Extend mode (10 Mbps per port over 250 m) is documented in notes for the 48P4S but not modelled - the port is still declared at 1G. ## Validation - `node scripts/import-devices.js --dry-run` over the batch: `Scanned 1 file(s), 6 spec(s) / Would add: 6 duplicates: 0 rejected: 0` - `node scripts/import-devices.js --apply`: added 6, no capability tokens dropped (all are engine tokens: l2-forward, vlan, lacp, stp, cli, http-server, poe-pse) - `node scripts/audit-port-types.js`: `No port contradicts its declared type.` (exit 0) - `node scripts/audit-devices.js`: `No hard defects.` - `node scripts/syntax-check.js`: 333 files, all OK - `node --test test/importer-roundtrip.test.js test/import-devices-capabilities.test.js test/fuzz-importer.test.js`: 26 passed, 0 failed - `node --test test/catalogue-hygiene.test.js`: 25 passed, 0 failed - Engine load check: 48P4S resolves 53 interfaces, 48 with 36 W `poeOut`, port 49 cages as SFP/optical.
catalogue(planet): six GS-4210 L2/L4 managed gigabit switches
Some checks failed
CI / Classify the change (pull_request) Successful in 23s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 31s
CI / Docker build (pull_request) Failing after 10m13s
CI / Server smoke test (pull_request) Failing after 10m14s
CI / Test (Node 22) (pull_request) Failing after 10m19s
CI / Test (Node 20) (pull_request) Failing after 10m21s
CI / Test (Node 18) (pull_request) Failing after 10m25s
b28ca2e69b
Adds the PLANET GS-4210 managed access family with datasheet-sourced
port counts, PoE budgets and per-port wattage:

  GS-4210-8P2S   8x1G PoE+ (30 W) + 2x 100/1000X SFP, 140 W, 20 Gbps
  GS-4210-16P2S  16x1G PoE+ (30 W) + 2x 100/1000X SFP, 220 W, 36 Gbps
  GS-4210-24P2S  24x1G PoE+ (30 W) + 2x 100/1000/2500X SFP, 370 W, 58 Gbps (hw v4)
  GS-4210-24T4S  24x1G + 4x 1000/2500X SFP, no PoE, 68 Gbps (hw v2)
  GS-4210-48T4S  48x1G + 4x 100/1000X SFP, no PoE, 104 Gbps (hw v3)
  GS-4210-48P4S  48x1G PoE+ (36 W) + 4x 100/1000X SFP, 600 W, 104 Gbps

Every field carries the datasheet or product-page URL it came from, and
each entry's notes name the hardware revision it models, because PLANET
restates the same model number with different PoE budgets and switch
fabric across revisions (e.g. 24P2S: 300 W QIG / 370 W v4).
Author
Owner

This branch conflicts with main, so it was not rebased:

devices/planet/gs-4210-16p2s.json devices/planet/gs-4210-24p2s.json devices/planet/gs-4210-24t4s.json devices/planet/gs-4210-48p4s.json devices/planet/gs-4210-48t4s.json devices/planet/gs-4210-8p2s.json research/incoming/devices-planet-gs4210.json 

A conflict is a decision about behaviour rather than a mechanical fix, so this is left for a human. Rebasing by hand and force-pushing will let the automation pick it up again.

This branch conflicts with `main`, so it was not rebased: ``` devices/planet/gs-4210-16p2s.json devices/planet/gs-4210-24p2s.json devices/planet/gs-4210-24t4s.json devices/planet/gs-4210-48p4s.json devices/planet/gs-4210-48t4s.json devices/planet/gs-4210-8p2s.json research/incoming/devices-planet-gs4210.json ``` A conflict is a decision about behaviour rather than a mechanical fix, so this is left for a human. Rebasing by hand and force-pushing will let the automation pick it up again.
Author
Owner

Needs Jay's decision — the three add/add files are two different devices, not two copies.

#1590 landed PLANET GS-4210 files on main at 04:47; this PR adds its own versions of the same paths. The hardware facts disagree:

model main (f3bed6e3 / ddeb48fe / 045ae9f9) this PR (003758b7 / b49b2551 / 0450939c)
GS-4210-16P2S hw v3, 2x 100/1000/2500BASE-X SFP, 240 W budget, 42 Gbps 2x 100/1000BASE-X SFP, 220 W, 36 Gbps
GS-4210-24P2S (same class of difference) (same)
GS-4210-8P2S (same class of difference) (same)

One of the two datasheet readings is wrong and I cannot tell which from here — keeping either side silently ships a wrong port template. The PR also adds 3 models main does not have, so closing it outright would lose those. Decision: which dataset is authoritative for the three overlapping models (then the other 3 land cleanly), or split the new models into a fresh PR and close this one.

**Needs Jay's decision — the three add/add files are two different devices, not two copies.** #1590 landed PLANET GS-4210 files on `main` at 04:47; this PR adds its own versions of the same paths. The hardware facts disagree: | model | main (`f3bed6e3` / `ddeb48fe` / `045ae9f9`) | this PR (`003758b7` / `b49b2551` / `0450939c`) | |---|---|---| | GS-4210-16P2S | hw v3, 2x 100/1000/**2500**BASE-X SFP, 240 W budget, 42 Gbps | 2x 100/1000BASE-X SFP, 220 W, 36 Gbps | | GS-4210-24P2S | (same class of difference) | (same) | | GS-4210-8P2S | (same class of difference) | (same) | One of the two datasheet readings is wrong and I cannot tell which from here — keeping either side silently ships a wrong port template. The PR also adds 3 models `main` does not have, so closing it outright would lose those. Decision: which dataset is authoritative for the three overlapping models (then the other 3 land cleanly), or split the new models into a fresh PR and close this one.
Author
Owner

Narrowing the decision: only one of the three overlapping files actually disagrees.

I re-read both sides of each add/add file at main = 0c87c74c against this branch = b28ca2e6, and the picture is smaller than my note above:

model main this PR verdict
GS-4210-8P2S 140 W budget, 20 Gbps fabric, 2x 100/1000BASE-X SFP 140 W, 20 Gbps, 2x 100/1000BASE-X SFP agree — main wins, branch copy is redundant
GS-4210-24P2S 370 W, 58 Gbps, 2x 100/1000/2500BASE-X SFP, hw v4 370 W, 58 Gbps, 2x 100/1000/2500BASE-X SFP agree — main wins, branch copy is redundant
GS-4210-16P2S hw v3, 240 W, 42 Gbps, 2x 2500BASE-X SFP 220 W, 36 Gbps, 2x 100/1000BASE-X SFP genuinely different hardware reading

So the semantic question is now one model, not three: which datasheet is authoritative for the GS-4210-16P2S (240 W / 42 Gbps with 2.5G uplinks, or 220 W / 36 Gbps with 100/1000X uplinks). Main's 16P2S reads as a later hardware revision than the branch's, and its 24P2S copy already carries the matching "v4 vs v2/v3 revisions genuinely differ" note — so main is the more likely winner, but that is a hardware fact I cannot verify from here and will not guess.

The other three models this PR adds — gs-4210-24t4s, gs-4210-48p4s, gs-4210-48t4s — exist only here and are worth keeping.

Once the 16P2S figure is settled this becomes mechanical: keep main for the two agreeing files, add the three new models, and the rebase has nothing left to argue about. Also note all five CI jobs are red on b28ca2e6 (run 3243), and this Forgejo build returns 404 on the actions log API, so I could not read the failure reason — it will need a green run after the rebase regardless.

— surgery loop, 2026-11-16 07:23 run

**Narrowing the decision: only one of the three overlapping files actually disagrees.** I re-read both sides of each add/add file at `main` = `0c87c74c` against this branch = `b28ca2e6`, and the picture is smaller than my note above: | model | main | this PR | verdict | |---|---|---|---| | GS-4210-8P2S | 140 W budget, 20 Gbps fabric, 2x 100/1000BASE-X SFP | 140 W, 20 Gbps, 2x 100/1000BASE-X SFP | **agree** — main wins, branch copy is redundant | | GS-4210-24P2S | 370 W, 58 Gbps, 2x 100/1000/2500BASE-X SFP, hw v4 | 370 W, 58 Gbps, 2x 100/1000/2500BASE-X SFP | **agree** — main wins, branch copy is redundant | | GS-4210-16P2S | hw v3, 240 W, 42 Gbps, 2x **2500**BASE-X SFP | 220 W, 36 Gbps, 2x 100/1000BASE-X SFP | **genuinely different hardware reading** | So the semantic question is now one model, not three: which datasheet is authoritative for the **GS-4210-16P2S** (240 W / 42 Gbps with 2.5G uplinks, or 220 W / 36 Gbps with 100/1000X uplinks). Main's 16P2S reads as a later hardware revision than the branch's, and its 24P2S copy already carries the matching "v4 vs v2/v3 revisions genuinely differ" note — so main is the more likely winner, but that is a hardware fact I cannot verify from here and will not guess. The other three models this PR adds — `gs-4210-24t4s`, `gs-4210-48p4s`, `gs-4210-48t4s` — exist only here and are worth keeping. **Once the 16P2S figure is settled this becomes mechanical:** keep main for the two agreeing files, add the three new models, and the rebase has nothing left to argue about. Also note all five CI jobs are red on `b28ca2e6` (run 3243), and this Forgejo build returns 404 on the actions log API, so I could not read the failure reason — it will need a green run after the rebase regardless. — surgery loop, 2026-11-16 07:23 run
Author
Owner

Split: the three undisputed models now live in #1631. This PR keeps the only real question.

I ran the full suite on this branch's head (b28ca2e6) locally - 263 of 265 suites, 9,526 assertions, 2 failing suites:

  • doc-catalog-counts - genuinely stale generated docs: the registry headline is 2,109 / 321 switches, the catalogue now says 2,115 / 327. npm run regen:registry-doc fixes it. Not yours to decide.
  • static-surface - connect ECONNREFUSED 127.0.0.1:16989, a port collision on my machine, not this branch.

So nothing in the six device files is broken. The only thing standing between this branch and a clean rebase is still the GS-4210-16P2S reading (main: hw v3, 2x 100/1000/2500X SFP, 240 W, 42 Gbps - this branch: 100/1000X, 220 W, 36 Gbps).

What I did in this run: took the three models main does not have (24T4S, 48T4S, 48P4S) verbatim from this branch, regenerated docs/DEVICE_REGISTRY.md on top of current main (0571a52a), and opened #1631 so they are not held hostage by the 16P2S question. Locally green there: doc-catalog-counts 24/24, validation 22/22, hardware-invariants 26/26, audit-port-types, control bytes, syntax (343 files).

Once you say which 16P2S revision is authoritative, this PR reduces to: drop gs-4210-8p2s.json and gs-4210-24p2s.json (identical to main), keep or drop gs-4210-16p2s.json per your call, and regen:registry-doc. That is a mechanical rebase I can do in one run.

**Split: the three undisputed models now live in #1631.** This PR keeps the only real question. I ran the full suite on this branch's head (`b28ca2e6`) locally - 263 of 265 suites, 9,526 assertions, 2 failing suites: - `doc-catalog-counts` - genuinely stale generated docs: the registry headline is 2,109 / 321 switches, the catalogue now says 2,115 / 327. `npm run regen:registry-doc` fixes it. Not yours to decide. - `static-surface` - `connect ECONNREFUSED 127.0.0.1:16989`, a port collision on my machine, not this branch. So nothing in the six device files is broken. The only thing standing between this branch and a clean rebase is still the **GS-4210-16P2S** reading (main: hw v3, 2x 100/1000/2500X SFP, 240 W, 42 Gbps - this branch: 100/1000X, 220 W, 36 Gbps). What I did in this run: took the three models `main` does not have (`24T4S`, `48T4S`, `48P4S`) verbatim from this branch, regenerated `docs/DEVICE_REGISTRY.md` on top of current `main` (`0571a52a`), and opened **#1631** so they are not held hostage by the 16P2S question. Locally green there: doc-catalog-counts 24/24, validation 22/22, hardware-invariants 26/26, audit-port-types, control bytes, syntax (343 files). Once you say which 16P2S revision is authoritative, this PR reduces to: drop `gs-4210-8p2s.json` and `gs-4210-24p2s.json` (identical to main), keep or drop `gs-4210-16p2s.json` per your call, and `regen:registry-doc`. That is a mechanical rebase I can do in one run.
Author
Owner

Closing as superseded — all seven files in this PR are now on main, and the three PoE models here would regress what main already has.

Verified against origin/main at 1901308d:

file state on main
devices/planet/gs-4210-24t4s.json landed verbatim in 6a89c788 (PR #1631, split out of this branch)
devices/planet/gs-4210-48t4s.json landed verbatim in 6a89c788
devices/planet/gs-4210-48p4s.json landed verbatim in 6a89c788
devices/planet/gs-4210-8p2s.json already on main from ff2d4951 (PR #1590)
devices/planet/gs-4210-16p2s.json already on main from ff2d4951
devices/planet/gs-4210-24p2s.json already on main from ff2d4951
research/incoming/devices-planet-gs4210.json no loss — its six records are the same objects as the shipped device files, which already carry sourceUrl + notes

The three PoE models are not a straight duplicate: ff2d4951 (Jay's own PR, co-committed by jay) landed them in the richer schema — portTemplate with per-port speedCaps, 802.3af/802.3at capability tokens, and a fabric block — while this branch carries the older draft shape (bare numeric interface names, no portTemplate, no PoE standard tokens). Merging this PR over main would remove those, so a rebase here is not the answer.

The 16P2S question is already answered on main: ff2d4951 models hardware version 3 — 240 W, 42 Gbps, 2.5G SFP uplinks. This branch has hw v2 (220 W, 36 Gbps, 1000/1000X). If the v2 reading is the one you want, it is a one-file edit against main — reopen or file a fresh PR with just that change; I am not going to overwrite main's v3 numbers on my own judgement.

CI on head b28ca2e6 was red on all five functional jobs (Test 18/20/22, smoke, Docker) after ~10 min each, and mergeable=false at 41 commits behind — both moot now.

Branch catalogue/1116-0425 left in place, nothing rewritten. Jay — nothing needed here unless you want the 16P2S hw v2 figures instead.

**Closing as superseded — all seven files in this PR are now on `main`, and the three PoE models here would regress what main already has.** Verified against `origin/main` at `1901308d`: | file | state on main | |---|---| | `devices/planet/gs-4210-24t4s.json` | **landed verbatim** in `6a89c788` (PR #1631, split out of this branch) | | `devices/planet/gs-4210-48t4s.json` | **landed verbatim** in `6a89c788` | | `devices/planet/gs-4210-48p4s.json` | **landed verbatim** in `6a89c788` | | `devices/planet/gs-4210-8p2s.json` | already on main from `ff2d4951` (PR #1590) | | `devices/planet/gs-4210-16p2s.json` | already on main from `ff2d4951` | | `devices/planet/gs-4210-24p2s.json` | already on main from `ff2d4951` | | `research/incoming/devices-planet-gs4210.json` | no loss — its six records are the same objects as the shipped device files, which already carry `sourceUrl` + `notes` | The three PoE models are not a straight duplicate: `ff2d4951` (Jay's own PR, co-committed by jay) landed them in the richer schema — `portTemplate` with per-port `speedCaps`, `802.3af`/`802.3at` capability tokens, and a `fabric` block — while this branch carries the older draft shape (bare numeric interface names, no `portTemplate`, no PoE standard tokens). Merging this PR over main would remove those, so a rebase here is not the answer. **The 16P2S question is already answered on main:** `ff2d4951` models hardware version 3 — 240 W, 42 Gbps, 2.5G SFP uplinks. This branch has hw v2 (220 W, 36 Gbps, 1000/1000X). If the v2 reading is the one you want, it is a one-file edit against main — reopen or file a fresh PR with just that change; I am not going to overwrite main's v3 numbers on my own judgement. CI on head `b28ca2e6` was red on all five functional jobs (Test 18/20/22, smoke, Docker) after ~10 min each, and `mergeable=false` at 41 commits behind — both moot now. Branch `catalogue/1116-0425` left in place, nothing rewritten. Jay — nothing needed here unless you want the 16P2S hw v2 figures instead.
jaylouisw closed this pull request 2026-11-16 08:25:49 +00:00
Some checks failed
CI / Classify the change (pull_request) Successful in 23s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 31s
CI / Docker build (pull_request) Failing after 10m13s
CI / Server smoke test (pull_request) Failing after 10m14s
CI / Test (Node 22) (pull_request) Failing after 10m19s
CI / Test (Node 20) (pull_request) Failing after 10m21s
CI / Test (Node 18) (pull_request) Failing after 10m25s

Pull request closed

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!1603
No description provided.