catalogue(oceanprotect): X8100/X9100 management-plane census from the vendor white paper (#743) #1902

Merged
jaylouisw merged 2 commits from storage/743 into main 2026-11-17 22:05:53 +00:00
Owner

Fixes #743.

The request was parked on a front-end port count Huawei does not publish. The same array's management plane is published per model, in a document that is not behind the support portal at all: Huawei OceanProtect X Series Backup Appliance Technical White Paper, Issue 01 (2026-09-29), whose material page is https://e.huawei.com/en/documents/products/storage/eded1b265ecc4dfba2fbf8669b0b8213. Chapter 3 goes model by model, and both models in this request are in it.

What is modelled

X8100 X9100
vendor sentence "Each controller has two GE management/maintenance ports and one serial port" "Each controller enclosure has two management modules. Each management module has three GE ports (management/maintenance network ports), one USB port, and one serial port"
controllers per node 2 4
ports modelled 4 x 1GbE (A-/B-MGMT0, A-/B-MGMT1) 6 x 1GbE (MM0/MM1, two MGMT and one MAINT each)

They are two catalogue entries on purpose: 2 U / 2 controllers against 4 U / 4 controllers, which is what the earlier passes established and what .../x100-series states.

Frame and capacity facts are in notes, not modelled as ports: 12 hot-swappable interface-module slots and 36 NVMe disk slots for the X8100 (1-2 PiB usable per node, 86.1 x 447 x 920 mm, <= 45 kg), 28 shared interface-module slots for the X9100 (2-3 PiB usable per node, 175 x 447 x 865 mm, <= 89 kg), 4 / 8 controllers maximum per system, C13 or C19 feed. Serial and USB ports are in notes per docs/DEVICE_SCHEMA.md ("Serial consoles, USB, FXS"), and the connector is omitted rather than guessed as RJ45 - the vendor does not publish it.

What is still absent, and why that is not a gap in the entry

Every front-end port on these arrays comes from an interface module, and no reachable vendor book states a ports-per-module count. The white paper's section 2.3.2 gives a slot preference order (IOM 0 > IOM 1 > IOM 3 > IOM 4 > IOM 5 > IOM 2) where the 1.3.0 X6000/X8000 guide gave a restriction ("can be installed only in slots IOM 0, IOM 1, and IOM 4"), and 3.1.5-3.1.6 list the module kinds (8/16/32/64 Gbit/s FC, 10GE, 25GE, 40GE, 100GE, 100 Gbit/s RDMA scale-out, 12 Gbit/s SAS) with no counts. EDOC1100368229, EDOC1100368230, EDOC1100305240 and EDOC1100305243 all still answer HTTP 302 with X-ALB-HIT: req-hd-bl (measured 2026-11-17). So the entries model the build the vendor describes - node, no interface modules - rather than invent an IOM population.

research/oceanprotect-x-series-port-census.md gains sections 14-15 recording this pass, and section 15 carries the other request's negative result (see the comment on #742).

Gates

  • node scripts/syntax-check.js - 395 files, all OK
  • node scripts/audit-port-types.js - exit 0, no port contradicts its declared type
  • importer capability tripwire - 2 specs added, 0 rejected, 0 capability losses, 0 collisions
  • test/hardware-invariants.test.js - 26 passed, 0 failed (2,350 models)
  • test/catalogue-hygiene.test.js - 25 passed, 0 failed
  • test/doc-catalog-counts.test.js - 43 passed, 0 failed (regen:registry-doc run: 2,350 models, 29,245 sockets, 967 shapes)
  • test/spec-vs-description.test.js - 3 passed, 0 failed
Fixes #743. The request was parked on a front-end port count Huawei does not publish. The same array's **management plane is published per model**, in a document that is not behind the support portal at all: *Huawei OceanProtect X Series Backup Appliance Technical White Paper*, Issue 01 (2026-09-29), whose material page is `https://e.huawei.com/en/documents/products/storage/eded1b265ecc4dfba2fbf8669b0b8213`. Chapter 3 goes model by model, and both models in this request are in it. ## What is modelled | | X8100 | X9100 | |---|---|---| | vendor sentence | "Each controller has two GE management/maintenance ports and one serial port" | "Each controller enclosure has two management modules. Each management module has three GE ports (management/maintenance network ports), one USB port, and one serial port" | | controllers per node | 2 | 4 | | ports modelled | 4 x 1GbE (A-/B-MGMT0, A-/B-MGMT1) | 6 x 1GbE (MM0/MM1, two MGMT and one MAINT each) | They are two catalogue entries on purpose: 2 U / 2 controllers against 4 U / 4 controllers, which is what the earlier passes established and what `.../x100-series` states. Frame and capacity facts are in `notes`, not modelled as ports: 12 hot-swappable interface-module slots and 36 NVMe disk slots for the X8100 (1-2 PiB usable per node, 86.1 x 447 x 920 mm, <= 45 kg), 28 shared interface-module slots for the X9100 (2-3 PiB usable per node, 175 x 447 x 865 mm, <= 89 kg), 4 / 8 controllers maximum per system, C13 or C19 feed. Serial and USB ports are in `notes` per `docs/DEVICE_SCHEMA.md` ("Serial consoles, USB, FXS"), and the connector is omitted rather than guessed as RJ45 - the vendor does not publish it. ## What is still absent, and why that is not a gap in the entry Every front-end port on these arrays comes from an interface module, and no reachable vendor book states a ports-per-module count. The white paper's section 2.3.2 gives a slot **preference** order (`IOM 0 > IOM 1 > IOM 3 > IOM 4 > IOM 5 > IOM 2`) where the 1.3.0 X6000/X8000 guide gave a **restriction** ("can be installed only in slots IOM 0, IOM 1, and IOM 4"), and 3.1.5-3.1.6 list the module kinds (8/16/32/64 Gbit/s FC, 10GE, 25GE, 40GE, 100GE, 100 Gbit/s RDMA scale-out, 12 Gbit/s SAS) with no counts. `EDOC1100368229`, `EDOC1100368230`, `EDOC1100305240` and `EDOC1100305243` all still answer HTTP 302 with `X-ALB-HIT: req-hd-bl` (measured 2026-11-17). So the entries model the build the vendor describes - node, no interface modules - rather than invent an IOM population. `research/oceanprotect-x-series-port-census.md` gains sections 14-15 recording this pass, and section 15 carries the *other* request's negative result (see the comment on #742). ## Gates - `node scripts/syntax-check.js` - 395 files, all OK - `node scripts/audit-port-types.js` - exit 0, no port contradicts its declared type - importer capability tripwire - 2 specs added, 0 rejected, 0 capability losses, 0 collisions - `test/hardware-invariants.test.js` - 26 passed, 0 failed (2,350 models) - `test/catalogue-hygiene.test.js` - 25 passed, 0 failed - `test/doc-catalog-counts.test.js` - 43 passed, 0 failed (`regen:registry-doc` run: 2,350 models, 29,245 sockets, 967 shapes) - `test/spec-vs-description.test.js` - 3 passed, 0 failed
catalogue(oceanprotect): X8100/X9100 management-plane census from the vendor white paper (#743)
Some checks failed
CI / Classify the change (pull_request) Successful in 9s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 9s
CI / Server smoke test (pull_request) Successful in 7s
CI / Docker build (pull_request) Successful in 25s
CI / Test (Node 20) (pull_request) Failing after 45s
CI / Test (Node 22) (pull_request) Failing after 50s
CI / Test (Node 18) (pull_request) Failing after 55s
CI / Arm auto-merge (pull_request) Successful in 3s
b2d99ea7af
The pair was parked on a front-end port count Huawei does not publish, while the same array's
management plane is published per model in a document that is not on support.huawei.com at
all: OceanProtect X Series Backup Appliance Technical White Paper, Issue 01 (2026-09-29).

- X8100 (2 U, 2 controllers): 'Each controller has two GE management/maintenance ports and one
  serial port' -> 4 x 1GbE management ports; 12 hot-swappable interface-module slots, 36 NVMe
  slots, 1-2 PiB usable per node.
- X9100 (4 U, 4 controllers): 'Each controller enclosure has two management modules. Each
  management module has three GE ports' -> 6 x 1GbE; 28 shared interface-module slots, 2-3 PiB
  usable per node.

Both are modelled as the node with no interface modules fitted, because every front-end port on
these arrays comes from an IOM and no reachable vendor book states a ports-per-module count -
the white paper gives a slot *preference* order, not a rule, and EDOC1100368229/36230/305240
still answer 302 with X-ALB-HIT: req-hd-bl. Serial and USB ports are in notes per the schema;
the connector is omitted rather than guessed as RJ45.

research/oceanprotect-x-series-port-census.md sections 14-15 record this pass, including the
E8000 negative: the white paper that publishes this field for six X-series models contains
'E8000' zero times.
Author
Owner

Surgery pass: rebased onto main and fixed the red run; no content changed.

CI / Test was failing on one deterministic assertion, not on anything this PR argues:

test/device-aliases.test.js
not ok - no alias is empty, duplicated, self-referential, or another model name
  [oceanprotect-x8100: alias is its own model name ;
   oceanprotect-x8100: alias "OceanProtect X8100" is another device's model (huawei/oceanprotect-x8100) ; ...x9100 likewise]

The long alias reproduces the model field (model is the slug oceanprotect-x9100, the alias normalises to it), which the invariant rejects twice over. I dropped "OceanProtect X8100" / "OceanProtect X9100" and kept the short "X8100" / "X9100" forms, so nothing becomes less findable. That is the only change: no interface, note, port or research text was touched, and npm run regen:registry-doc produces no diff.

Verified in a scratch worktree, serial: device-aliases 13/0, hardware-invariants 26/0, catalogue-hygiene 25/0, missing-devices 66/0. Pushed as ddcd1e0d (rebase of b2d99ea7 onto 4881bce4, --force-with-lease).

If the census genuinely wants the long form searchable, the fix is the other way round — make model the display name as oceanprotect-x8000.json does — but that changes what the registry renders, so it is the author's call, not mine.

**Surgery pass: rebased onto main and fixed the red run; no content changed.** `CI / Test` was failing on one deterministic assertion, not on anything this PR argues: ``` test/device-aliases.test.js not ok - no alias is empty, duplicated, self-referential, or another model name [oceanprotect-x8100: alias is its own model name ; oceanprotect-x8100: alias "OceanProtect X8100" is another device's model (huawei/oceanprotect-x8100) ; ...x9100 likewise] ``` The long alias reproduces the model field (`model` is the slug `oceanprotect-x9100`, the alias normalises to it), which the invariant rejects twice over. I dropped `"OceanProtect X8100"` / `"OceanProtect X9100"` and kept the short `"X8100"` / `"X9100"` forms, so nothing becomes less findable. That is the only change: no interface, note, port or research text was touched, and `npm run regen:registry-doc` produces no diff. Verified in a scratch worktree, serial: `device-aliases` 13/0, `hardware-invariants` 26/0, `catalogue-hygiene` 25/0, `missing-devices` 66/0. Pushed as `ddcd1e0d` (rebase of `b2d99ea7` onto `4881bce4`, `--force-with-lease`). If the census genuinely wants the long form searchable, the fix is the other way round — make `model` the display name as `oceanprotect-x8000.json` does — but that changes what the registry renders, so it is the author's call, not mine.
jaylouisw force-pushed storage/743 from b2d99ea7af
Some checks failed
CI / Classify the change (pull_request) Successful in 9s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 9s
CI / Server smoke test (pull_request) Successful in 7s
CI / Docker build (pull_request) Successful in 25s
CI / Test (Node 20) (pull_request) Failing after 45s
CI / Test (Node 22) (pull_request) Failing after 50s
CI / Test (Node 18) (pull_request) Failing after 55s
CI / Arm auto-merge (pull_request) Successful in 3s
to ddcd1e0daf
All checks were successful
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 7s
CI / Classify the change (pull_request) Successful in 10s
CI / Server smoke test (pull_request) Successful in 9s
CI / Docker build (pull_request) Successful in 24s
CI / Test (Node 20) (pull_request) Successful in 49s
CI / Test (Node 18) (pull_request) Successful in 57s
CI / Test (Node 22) (pull_request) Successful in 54s
CI / Arm auto-merge (pull_request) Successful in 21s
2026-11-17 22:04:11 +00:00
Compare
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!1902
No description provided.