devices: NetEngine 8000 F8 in Huawei's own 6.4T bundle (#789) #1758

Closed
jaylouisw wants to merge 1 commit from catalogue/1117-0025 into main
Owner

Fixes #789

The NetEngine 8000 F8 is modellable because Huawei publishes a board census and quotes typical
power for a named board mix — the thing the earlier passes looked for on the marketing page and
did not find one document further down.

devices/huawei/netengine-8000-f8.json — model NetEngine 8000 F8 (4x 8x100GE, 4x 72x10GE),
320 ports:

  • 32x 100GE/50GE/40GE QSFP28 (CR8DE8NB00F0, LPUC-800-CM, 8 ports, slots 1-4)
  • 288x 10GE/GE SFP+ (CR8DERXFMSF0, LPUC-800-CM, 72 ports, MACsec, slots 5-8)
  • 6.08 Tbit/s of ports inside the 28.86 Tbit/s 6.4T-bundle fabric, 2362 W typical, 13U, 8 LPU slots

Why that mix is not a guess: the chassis table states "Typical power consumption 2362 W
(configured with SRU 6T4-B and four 8x100GE ports and four 72x10GE ports)". Which slots hold which
board is the one field the source does not state — the spec says so in notes, and any LPUC-800-CM
board may sit in any of LPU slots 1-8 with an SRU-6T4-B.

Port names follow the guide's own rule (§3.5.1.4): slot ID/PIC ID/port number, PIC IDs from 0,
LPUC PIC ID 0, ports from 0 left to right.

Sources (in sourceUrl + notes):

research/huawei-netengine-8000-f8-boards.md records the board table (8x100GE, 4x100GE+40x10GE,
72x10GE/GE, LPUC-100 72x GE, LPUC-400-CM, LPUT-800-CM + PICs), the slot map and the RTU caveat
(high-speed modes are capped at 100 Mbit/s until the licence loads), so the remaining F8 variants are
one spec file each rather than another research pass.

Gates

  • node scripts/syntax-check.js — 367 files, all OK
  • node scripts/audit-port-types.js — "No port contradicts its declared type."
  • scripts/import-devices.js (PF_INCOMING/PF_DEVICES sandbox) — "Scanned 1 file(s), 1 spec(s) / Would add: 1 duplicates: 0 rejected: 0", no capability warnings
  • test/hardware-invariants.test.js — 26 passed, 0 failed (2,340 models)
  • test/catalogue-hygiene.test.js — 25 passed, 0 failed
  • test/faceplate-catalogue.test.js — 14 passed, 0 failed
  • scripts/regen-registry-doc.js — registry doc + counts regenerated, never hand-edited
Fixes #789 The NetEngine 8000 F8 is modellable because Huawei publishes a **board census** and quotes typical power for a **named board mix** — the thing the earlier passes looked for on the marketing page and did not find one document further down. **`devices/huawei/netengine-8000-f8.json`** — model `NetEngine 8000 F8 (4x 8x100GE, 4x 72x10GE)`, 320 ports: - 32x 100GE/50GE/40GE QSFP28 (`CR8DE8NB00F0`, LPUC-800-CM, 8 ports, slots 1-4) - 288x 10GE/GE SFP+ (`CR8DERXFMSF0`, LPUC-800-CM, 72 ports, MACsec, slots 5-8) - 6.08 Tbit/s of ports inside the 28.86 Tbit/s 6.4T-bundle fabric, 2362 W typical, 13U, 8 LPU slots **Why that mix is not a guess:** the chassis table states "Typical power consumption 2362 W (configured with SRU 6T4-B and four 8x100GE ports and four 72x10GE ports)". Which slots hold which board is the one field the source does not state — the spec says so in `notes`, and any LPUC-800-CM board may sit in any of LPU slots 1-8 with an SRU-6T4-B. Port names follow the guide's own rule (§3.5.1.4): `slot ID/PIC ID/port number`, PIC IDs from 0, LPUC PIC ID 0, ports from 0 left to right. **Sources** (in `sourceUrl` + `notes`): - https://support.huawei.com/enterprise/en/doc/EDOC1100280456 — hardware guide, Issue 06 (2026-06-30); slots 1-8 LPU / 9-10 SRU / 11-12 MPU, board chapter, port numbering, RTU licence limits - https://e.huawei.com/marketingcloud/pep/asset/20000001/Material/e679822234f94651b2c2153c21a6621b/M2T1A669N893162565584830493/NetEngine%208000%20F8%20Universal%20Service%20Router%20Datasheet.pdf — 2022 datasheet; its 1680 W / 2 Tbps / "24 x 100GE; 320 x GE" figures are the older PIC-based bundle and are recorded as such rather than silently preferred - https://e.huawei.com/en/products/routers/ne8000 — the F8 column (8 line cards, 28.86 Tbit/s, 121.6 kg) **`research/huawei-netengine-8000-f8-boards.md`** records the board table (8x100GE, 4x100GE+40x10GE, 72x10GE/GE, LPUC-100 72x GE, LPUC-400-CM, LPUT-800-CM + PICs), the slot map and the RTU caveat (high-speed modes are capped at 100 Mbit/s until the licence loads), so the remaining F8 variants are one spec file each rather than another research pass. **Gates** - `node scripts/syntax-check.js` — 367 files, all OK - `node scripts/audit-port-types.js` — "No port contradicts its declared type." - `scripts/import-devices.js` (PF_INCOMING/PF_DEVICES sandbox) — "Scanned 1 file(s), 1 spec(s) / Would add: 1 duplicates: 0 rejected: 0", no capability warnings - `test/hardware-invariants.test.js` — 26 passed, 0 failed (2,340 models) - `test/catalogue-hygiene.test.js` — 25 passed, 0 failed - `test/faceplate-catalogue.test.js` — 14 passed, 0 failed - `scripts/regen-registry-doc.js` — registry doc + counts regenerated, never hand-edited
devices: NetEngine 8000 F8 modelled in Huawei's own 6.4T bundle (#789)
All checks were successful
CI / Classify the change (pull_request) Successful in 16s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 20s
CI / Server smoke test (pull_request) Successful in 19s
CI / Test (Node 20) (pull_request) Successful in 58s
CI / Test (Node 18) (pull_request) Successful in 1m2s
CI / Docker build (pull_request) Successful in 29s
CI / Test (Node 22) (pull_request) Successful in 56s
CI / Arm auto-merge (pull_request) Successful in 5s
c7c64d7d78
jaylouisw force-pushed catalogue/1117-0025 from c7c64d7d78
All checks were successful
CI / Classify the change (pull_request) Successful in 16s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 20s
CI / Server smoke test (pull_request) Successful in 19s
CI / Test (Node 20) (pull_request) Successful in 58s
CI / Test (Node 18) (pull_request) Successful in 1m2s
CI / Docker build (pull_request) Successful in 29s
CI / Test (Node 22) (pull_request) Successful in 56s
CI / Arm auto-merge (pull_request) Successful in 5s
to 0a65ccf819
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 18s
CI / Server smoke test (pull_request) Successful in 19s
CI / Test (Node 20) (pull_request) Successful in 57s
CI / Test (Node 18) (pull_request) Successful in 1m1s
CI / Docker build (pull_request) Successful in 31s
CI / Test (Node 22) (pull_request) Successful in 59s
CI / Arm auto-merge (pull_request) Successful in 12s
2026-11-17 00:38:59 +00:00
Compare
jaylouisw force-pushed catalogue/1117-0025 from 0a65ccf819
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 18s
CI / Server smoke test (pull_request) Successful in 19s
CI / Test (Node 20) (pull_request) Successful in 57s
CI / Test (Node 18) (pull_request) Successful in 1m1s
CI / Docker build (pull_request) Successful in 31s
CI / Test (Node 22) (pull_request) Successful in 59s
CI / Arm auto-merge (pull_request) Successful in 12s
to 62fe352840
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 16s
CI / Server smoke test (pull_request) Successful in 10s
CI / Test (Node 20) (pull_request) Successful in 48s
CI / Docker build (pull_request) Successful in 20s
CI / Test (Node 18) (pull_request) Successful in 56s
CI / Test (Node 22) (pull_request) Successful in 53s
CI / Arm auto-merge (pull_request) Successful in 1m35s
2026-11-17 00:49:55 +00:00
Compare
Author
Owner

Green on all eight checks, but it cannot be merged mechanically — and the reason is a modelling decision, not a conflict I should resolve. Needs Jay's decision.

State

  • Head 62fe3528: CI / Test (Node 18/20/22), Server smoke test, Docker build, Arm auto-merge — all success. Your only problem is mergeable=false, and that is real here, not an uncomputed value.
  • Merge base f520c044; main is at 18f25c1a (9 ahead).
  • Add/add conflict on devices/huawei/netengine-8000-f8.json. #1743 (same issue, #789) was opened 22:41, merged at 00:54 as 12a9afac; this PR was opened 00:37, seventeen minutes before that merge landed and after the point where it could see it. Both add the same path. Issue #789 is now closed by 12a9afac.

Why I did not rebase it

A rebase would not remove the conflict, it would merely turn it into "delete main's file, add mine" — i.e. I would be choosing a port census. That is yours.

main 12a9afac (#1743) this PR
model string netengine-8000-f8 NetEngine 8000 F8 (4x 8x100GE, 4x 72x10GE)
population 8 slots x 8 ports = 64 x 100GE, all slots 100GE cards (100GE1/0/0 … 100GE8/3/1) 4 slots x 8 = 32 x 100GE + 4 slots x 72 = 288 x 10GE SFP+ = 320 ports (100GE1/0/0 … 10GE8/0/71)
aliases NE8000 F8, NetEngine 8000E F8 adds CR8PBKPF8DC0, LPUC-800-CM
extras — research/huawei-netengine-8000-f8-boards.md (77 lines), docs/CONNECTORS.md, README, docs/wiki/Home.md, scripts/build-wiki.js

Both cite the same document family (datasheet V800R026C00); the difference is which sentence is the model of record — #1743 took the maximum interface configuration (every slot with a 100GE PIC), this PR took the 6.4T bundle board mix (CR8DE8NB00F0 x4 + CR8DERXFMSF0 x4). Neither is invented; they are two populations of one chassis.

Decision needed

  1. One canonical census — replace main's file with this PR's 320-port mix (rebase, then force-push; the add/add goes away because the content becomes a modification).
  2. Both, as variants — the repo already does this (netengine-8000-m1d.json vs -m1d-b.json, ar8140-12g10xg vs -12g10xg- variant, f1a vs f1a-c), so this PR could land as e.g. netengine-8000-f8-6.4t.json. I would need you to name the suffix rule; inventing it is how the alias tests start failing.
  3. Close as superseded — #789 is answered by 12a9afac and this is the same request twice.

Secondary, for whoever wins: LPUC-800-CM is a line-card part number, not a device alias; on main it lives in notes. Alias lists are asserted unique across the catalogue, so shipping it as an alias is a trap for the next device that uses that PIC.

Verified: both JSONs parsed; port counts summed from interfaces[] (64 vs 320); git merge-base --is-ancestor 12a9afac origin/main → true; GET /pulls/1758/files → 7 files, devices/huawei/netengine-8000-f8.json added at +3260. I have not pushed to or rebased this branch.

— surgery watchdog

**Green on all eight checks, but it cannot be merged mechanically — and the reason is a modelling decision, not a conflict I should resolve. Needs Jay's decision.** ### State - Head `62fe3528`: `CI / Test (Node 18/20/22)`, `Server smoke test`, `Docker build`, `Arm auto-merge` — all **success**. Your only problem is `mergeable=false`, and that is real here, not an uncomputed value. - Merge base `f520c044`; main is at `18f25c1a` (9 ahead). - **Add/add conflict on `devices/huawei/netengine-8000-f8.json`.** #1743 (same issue, #789) was opened 22:41, merged at **00:54 as `12a9afac`**; this PR was opened 00:37, seventeen minutes before that merge landed and after the point where it could see it. Both add the same path. Issue #789 is now closed by `12a9afac`. ### Why I did not rebase it A rebase would not remove the conflict, it would merely turn it into *"delete main's file, add mine"* — i.e. I would be choosing a port census. That is yours. | | main `12a9afac` (#1743) | this PR | |---|---|---| | model string | `netengine-8000-f8` | `NetEngine 8000 F8 (4x 8x100GE, 4x 72x10GE)` | | population | 8 slots x 8 ports = **64 x 100GE**, all slots 100GE cards (`100GE1/0/0` … `100GE8/3/1`) | 4 slots x 8 = **32 x 100GE** + 4 slots x 72 = **288 x 10GE SFP+** = **320 ports** (`100GE1/0/0` … `10GE8/0/71`) | | aliases | `NE8000 F8`, `NetEngine 8000E F8` | adds `CR8PBKPF8DC0`, `LPUC-800-CM` | | extras | — | `research/huawei-netengine-8000-f8-boards.md` (77 lines), `docs/CONNECTORS.md`, README, `docs/wiki/Home.md`, `scripts/build-wiki.js` | Both cite the same document family (datasheet `V800R026C00`); the difference is which sentence is the model of record — #1743 took the **maximum interface configuration** (every slot with a 100GE PIC), this PR took the **6.4T bundle board mix** (`CR8DE8NB00F0` x4 + `CR8DERXFMSF0` x4). Neither is invented; they are two populations of one chassis. ### Decision needed 1. **One canonical census** — replace main's file with this PR's 320-port mix (rebase, then force-push; the add/add goes away because the content becomes a modification). 2. **Both, as variants** — the repo already does this (`netengine-8000-m1d.json` vs `-m1d-b.json`, `ar8140-12g10xg` vs `-12g10xg-` variant, `f1a` vs `f1a-c`), so this PR could land as e.g. `netengine-8000-f8-6.4t.json`. I would need you to name the suffix rule; inventing it is how the alias tests start failing. 3. **Close as superseded** — #789 is answered by `12a9afac` and this is the same request twice. Secondary, for whoever wins: `LPUC-800-CM` is a *line-card* part number, not a device alias; on main it lives in `notes`. Alias lists are asserted unique across the catalogue, so shipping it as an alias is a trap for the next device that uses that PIC. Verified: both JSONs parsed; port counts summed from `interfaces[]` (64 vs 320); `git merge-base --is-ancestor 12a9afac origin/main` → true; `GET /pulls/1758/files` → 7 files, `devices/huawei/netengine-8000-f8.json` added at +3260. I have not pushed to or rebased this branch. — surgery watchdog
Author
Owner

Follow-up from the surgery watchdog — the conflict is narrower than "two F8s", and here is the exact shape of it.

I diffed the two versions of devices/huawei/netengine-8000-f8.json as they exist today:

main (from merged #1743) this PR
blob 59769c3f, 16,970 B f991e67a, 60,366 B
model netengine-8000-f8 NetEngine 8000 F8 (4x 8x100GE, 4x 72x10GE)
ports 32x CR8DE2NBMSF0 = 64x 100GE 32x 100GE (slots 1-4) + 288x 10GE (slots 5-8)
capacity claim 6.4 Tbit/s 6.08 Tbit/s of port capacity

So this is not a duplicated chassis, it is a collision on model / slug: both versions are the same physical 13U box with the same 28.86 Tbit/s fabric, and each encodes one legitimate line-card mix. Only one can own netengine-8000-f8.

The decision I cannot make for you is a catalogue rule, not a merge:

  1. Variant slugs — keep main's all-100GE as netengine-8000-f8 and land this as e.g. netengine-8000-f8-6.4t-bundle. Both survive, main stays green, but it sets a precedent that per-order line-card mixes each get their own slug — which then applies to every modular chassis we carry, not just this one.
  2. Replace — main's entry becomes the bundle mix and the all-100GE claim is dropped. One slug, one config, but we lose the "vendor maximum 100GE" build that #1743 was written to express.
  3. Fold into one entry — widen main's interfaces to cover both slot mixes. Cheapest for the registry, but the entry then describes a box that does not ship as configured, and interfaces is what the reachability/socket counts are derived from.

Verified before posting: head 62fe3528 is green on all eight checks; the PR is 15 commits behind main and the only conflicting path is this one JSON file (add/add) — regenerated docs and the registry surfaces are not in conflict.

**Follow-up from the surgery watchdog — the conflict is narrower than "two F8s", and here is the exact shape of it.** I diffed the two versions of `devices/huawei/netengine-8000-f8.json` as they exist today: | | main (from merged #1743) | this PR | |---|---|---| | blob | `59769c3f`, 16,970 B | `f991e67a`, 60,366 B | | model | `netengine-8000-f8` | `NetEngine 8000 F8 (4x 8x100GE, 4x 72x10GE)` | | ports | 32x CR8DE2NBMSF0 = **64x 100GE** | 32x 100GE (slots 1-4) + **288x 10GE** (slots 5-8) | | capacity claim | 6.4 Tbit/s | 6.08 Tbit/s of port capacity | So this is not a duplicated *chassis*, it is a **collision on `model` / slug**: both versions are the same physical 13U box with the same 28.86 Tbit/s fabric, and each encodes one legitimate line-card mix. Only one can own `netengine-8000-f8`. The decision I cannot make for you is a catalogue rule, not a merge: 1. **Variant slugs** — keep main's all-100GE as `netengine-8000-f8` and land this as e.g. `netengine-8000-f8-6.4t-bundle`. Both survive, main stays green, but it sets a precedent that per-order line-card mixes each get their own slug — which then applies to every modular chassis we carry, not just this one. 2. **Replace** — main's entry becomes the bundle mix and the all-100GE claim is dropped. One slug, one config, but we lose the "vendor maximum 100GE" build that #1743 was written to express. 3. **Fold into one entry** — widen main's `interfaces` to cover both slot mixes. Cheapest for the registry, but the entry then describes a box that does not ship as configured, and `interfaces` is what the reachability/socket counts are derived from. Verified before posting: head `62fe3528` is green on all eight checks; the PR is 15 commits behind main and the only conflicting path is this one JSON file (add/add) — regenerated docs and the registry surfaces are not in conflict.
Author
Owner

Closing as superseded — main already answers #789, and this entry can only land by replacing the one that is already there.

Verified this run on 64fb915d:

  • The issue this PR closes is closed. #789 [Device]: NetEngine 8000 F8 went closed at 2026-11-17T00:54:11Z, fixed by #1743 / 12a9afac (huawei: NetEngine 8000 F8, 64x 100GE from the vendor's own maximum configuration). That commit created devices/huawei/netengine-8000-f8.json on main.
  • So this is an add/add on the same path, not a stale-base conflict. git merge-tree against 64fb915d gives three stages for devices/huawei/netengine-8000-f8.json — main 59769c3f (16,970 B) vs this branch f991e67a — plus regenerated-doc conflicts in README.md, docs/CONNECTORS.md, docs/DEVICE_REGISTRY.md, docs/wiki/Home.md and one line in scripts/build-wiki.js. Only one commit on main has ever touched that device file, so there is nothing to rebase onto: the two entries are competing descriptions of one 13U chassis.
  • The two descriptions are not compatible as two catalogue rows. Main models the maximum 100GE build (32 PICs × CR8DE2NBMSF0 = 64 QSFP28 at 100GE, model netengine-8000-f8, matching the file path). This branch models a mixed 6.4T bundle (4× 8×100GE + 4× 72×10GE, 32 QSFP28 + 288 SFP+), with model set to NetEngine 8000 F8 (4x 8x100GE, 4x 72x10GE) — a display label in a field that is a slug everywhere else in the registry. Landing both would give the catalogue two spellings and two port counts for one box, which is exactly the failure #60 was raised about.
  • Your eight checks are green on 62fe3528 (Test 18/20/22, server smoke, docker build, rebase, auto-merge) — the branch itself is sound. Nothing is being thrown away: the commit is still reachable at refs/pull/1758/head.

If the bundle build is worth having, it needs its own identity — e.g. devices/huawei/netengine-8000-f8-6t4.json with model: "netengine-8000-f8-6t4" and a label carrying the 4×100GE + 4×72×10GE description — and that is a catalogue-shape decision for Jay, not something I will bake in. Reopen or re-cut against that path and it will merge cleanly.

Closed by the surgery watchdog (cron) under the superseded-PR rule: branch's issue already fixed on main, cite the commit, close.

**Closing as superseded — main already answers #789, and this entry can only land by replacing the one that is already there.** Verified this run on `64fb915d`: - **The issue this PR closes is closed.** #789 `[Device]: NetEngine 8000 F8` went `closed` at 2026-11-17T00:54:11Z, fixed by **#1743 / `12a9afac`** (`huawei: NetEngine 8000 F8, 64x 100GE from the vendor's own maximum configuration`). That commit created `devices/huawei/netengine-8000-f8.json` on main. - **So this is an add/add on the same path, not a stale-base conflict.** `git merge-tree` against `64fb915d` gives three stages for `devices/huawei/netengine-8000-f8.json` — main `59769c3f` (16,970 B) vs this branch `f991e67a` — plus regenerated-doc conflicts in `README.md`, `docs/CONNECTORS.md`, `docs/DEVICE_REGISTRY.md`, `docs/wiki/Home.md` and one line in `scripts/build-wiki.js`. Only one commit on main has ever touched that device file, so there is nothing to rebase *onto*: the two entries are competing descriptions of one 13U chassis. - **The two descriptions are not compatible as two catalogue rows.** Main models the maximum 100GE build (32 PICs × CR8DE2NBMSF0 = 64 QSFP28 at 100GE, model `netengine-8000-f8`, matching the file path). This branch models a mixed 6.4T bundle (4× 8×100GE + 4× 72×10GE, 32 QSFP28 + 288 SFP+), with `model` set to `NetEngine 8000 F8 (4x 8x100GE, 4x 72x10GE)` — a display label in a field that is a slug everywhere else in the registry. Landing both would give the catalogue two spellings and two port counts for one box, which is exactly the failure #60 was raised about. - Your eight checks are green on `62fe3528` (`Test` 18/20/22, server smoke, docker build, rebase, auto-merge) — the branch itself is sound. Nothing is being thrown away: the commit is still reachable at `refs/pull/1758/head`. **If the bundle build is worth having**, it needs its own identity — e.g. `devices/huawei/netengine-8000-f8-6t4.json` with `model: "netengine-8000-f8-6t4"` and a `label` carrying the 4×100GE + 4×72×10GE description — and that is a catalogue-shape decision for Jay, not something I will bake in. Reopen or re-cut against that path and it will merge cleanly. Closed by the surgery watchdog (cron) under the superseded-PR rule: branch's issue already fixed on main, cite the commit, close.
jaylouisw closed this pull request 2026-11-17 03:40:23 +00:00
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 16s
CI / Server smoke test (pull_request) Successful in 10s
CI / Test (Node 20) (pull_request) Successful in 48s
CI / Docker build (pull_request) Successful in 20s
CI / Test (Node 18) (pull_request) Successful in 56s
CI / Test (Node 22) (pull_request) Successful in 53s
CI / Arm auto-merge (pull_request) Successful in 1m35s

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!1758
No description provided.