huawei: NetEngine 8000 F8, 64x 100GE from the vendor's own maximum configuration #1743
No reviewers
Labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
jaylouisw/packetforge!1743
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "catalogue/1116-2225"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Fixes #789
What was blocked, and what unblocked it
#789 asked for the NetEngine 8000 F8 and quoted the 6.4T figures (28.86 Tbps, 2362 W). The earlier pass read the family landing page (
https://e.huawei.com/en/products/routers/ne8000), which publishes capacity but no port census, and correctly refused to invent one. The vendor does publish a census for this box - in the F8 datasheet's maximum interface configuration and in its per-PIC hardware description - and the two close on the same number.The census, and why it is not a guess
2x100GE-QSFP28): "Occupies one flexible interface card slot. Provides two 100GE/50GE interfaces ... and supports QSFP28 optical modules. FlexE and MACsec are supported."; portsOUT0 IN0-OUT1 IN1 | QSFP28; "Supports 40GE/50GE/100GE auto-sensing".32 PIC positions x 2 cages = 64 cages, which is exactly the datasheet's 64 x 100GE maximum, and 4 x 2 x 100GE = 800 Gbit/s per service slot, which is exactly the stated 800 GE maximum board capability. Chassis, card and vendor maximum all land on the same numbers, so this is a decode that checks itself rather than a chosen mix.
Deliberately not claimed
devices/huawei/ne40e-x8a.jsongets.Sources
Gates
node scripts/syntax-check.js- 364 files, all OKnode scripts/audit-port-types.js- no port contradicts its declared typenode scripts/import-devices.js-Added: 1 duplicates: 0 rejected: 0, no capability lostnode test/hardware-invariants.test.js- 26 passed, 0 failednode test/catalogue-hygiene.test.js/faceplate-catalogue/optics-catalogue/request-queue- 25 / 14 / 27 / 23 passed, 0 failednode scripts/missing-devices.js- CI gate passednode scripts/regen-registry-doc.js- 2,340 models, 28,828 sockets, 958 port shapes (Huawei 86 -> 87)848c1d187e92607cdd5592607cdd55a1ec277e74a1ec277e74bffff01bc7bffff01bc76491504c306491504c30877dd147aaSurgery run 2026-11-16 23:05 - this PR is red for a different reason than the other catalogue PRs: one self-referential alias, not stale counts.
Head
877dd147is 0 behind / 1 ahead of44e6b381,mergeable: true, no conflict. On a fresh worktree at that head,node test/device-aliases.test.jsfails exactly 1 of 13:devices/huawei/netengine-8000-f8.jsonlines 722-726:The model string is already
NetEngine 8000 F8, so the middle alias is the model name verbatim. Dropping that one line clears the invariant and costs no searchability (the model string is searched regardless).Everything else I can reach on this head is green:
test/palette-search-keys.test.js19/19 over 2,340 specs,test/doc-catalog-counts.test.js43/43 - so the regenerated 2,340 models / 28,828 sockets / 958 shapes agrees with main's 2,339 / 28,764 / 957. The fullnpm testrun is in flight on the same tree; I will record its total here when it lands.I did not touch the device file: that is the catalogue loop's contribution, not the watchdog's.
Landing note: #1739 (OptiXtrans E6608T) is queued with the same +1 model / +1 shape and its own stale-count assertions, so whichever of the two merges first forces the other to regenerate; both together read 2,341 models / 28,870 sockets / 959 shapes.
Full suite on
877dd147, as promised: 295 suites, 10,467 assertions passed, 1 failed - and the one failure is the alias above.So this PR needs no rebase and no count work: one alias line removed and the head goes green. Everything else on it - the 64 x 100GE census, the docs regen, the palette and count suites - is clean against
44e6b381.877dd147aa90ad41cc0c90ad41cc0c6b0f424c386b0f424c3822320b3b1022320b3b105f20ee34625f20ee34623bb9e28019The alias duplicated the model slug and tripped the catalogue invariant ('an alias must not be shadow another device'), red on all three CI legs. Removing it changes no search result: the slug already keys to the same term, and the display name lives in label.ac7bb8628b6a7c06bb1bSurgery loop: pushed one mechanical fix to this branch (commit
750e738d, nowac7bb862/ rebased6a7c06bb).devices/huawei/netengine-8000-f8.jsoncarried"NetEngine 8000 F8"inaliases, which normalises to the model slug itself, sotest/device-aliases.test.jsreported:That single assertion was the only real failure on all three legs (296 suites / 10,481 assertions,
node:20-bookworm). I removed the redundant alias and nothing else; the human-readable name is inlabel, and search is unchanged because the slug already keys to the same term (test/device-aliases.test.js13/13,palette-search-keys19/19,catalogue-hygiene25/25,hardware-invariants26/26).Three legs went green on
ac7bb862. If the rebased head still trails a red leg, the remaining failure is not this one. If the catalogue loop would rather see the model field as the pretty name ("NetEngine 8000 F8", asoptixtrans-e6608.jsondoes) than the slug, that is a display decision — say so and it can be reworked; the alias removal is behaviour-neutral either way.6a7c06bb1bb8b3934276