[Device]: Huawei OceanProtect X8100/X9100 Backup Storage #743

Closed
opened 2026-11-08 06:19:54 +00:00 by jaylouisw · 11 comments
Owner

Vendor

Huawei

Exact model name

Huawei OceanProtect X8100/X9100 Backup Storage

Category

NAS

Ports and interfaces

  • Front-end: 16/32/64 Gbit/s Fibre Channel
  • Front-end: 10/25/40/100GE
  • Protocols: NFS, SMB/CIFS, FC, iSCSI, S3, FTP/SFTP, NDMP
  • X8100: 2 controllers/node; X9100: 4 controllers/node (up to 4/8 controllers)
  • Usable capacity 1-6 PiB; NVMe SSD 7.68-122 TB

PoE budget (watts)

None

Datasheet or product page URL

https://e.huawei.com/en/products/storage/oceanprotect/x100-series

What do you want to model with it?

High-end all-flash backup storage (PBBA) as a backup target / integrated appliance on the network.

Capabilities

backup storage (NFS/SMB/iSCSI/S3/FC target), dedupe/compression

Before submitting

  • I searched the catalog and existing issues, and this model is not there yet.
  • The port list above comes from a datasheet, not from memory or another wiki.
### Vendor Huawei ### Exact model name Huawei OceanProtect X8100/X9100 Backup Storage ### Category NAS ### Ports and interfaces - Front-end: 16/32/64 Gbit/s Fibre Channel - Front-end: 10/25/40/100GE - Protocols: NFS, SMB/CIFS, FC, iSCSI, S3, FTP/SFTP, NDMP - X8100: 2 controllers/node; X9100: 4 controllers/node (up to 4/8 controllers) - Usable capacity 1-6 PiB; NVMe SSD 7.68-122 TB ### PoE budget (watts) None ### Datasheet or product page URL https://e.huawei.com/en/products/storage/oceanprotect/x100-series ### What do you want to model with it? High-end all-flash backup storage (PBBA) as a backup target / integrated appliance on the network. ### Capabilities backup storage (NFS/SMB/iSCSI/S3/FC target), dedupe/compression ### Before submitting - [x] I searched the catalog and existing issues, and this model is not there yet. - [x] The port list above comes from a datasheet, not from memory or another wiki.
jaylouisw referenced this issue from a commit 2026-11-08 06:43:31 +00:00
Author
Owner

Blocked on a port count, not on research effort.

Huawei's public page for the OceanProtect X8100/X9100 gives the port types - front-end 10/25/40/100GE and 16/32/64 Gbit/s Fibre Channel, with 2 or 4 controllers per node and up to 8 in a system - but not how many front-end ports a node has. That number depends on the interface-card configuration, and the configuration-specific tables live on support.huawei.com, which is behind a login wall I cannot pass (plain fetch returns HTTP 422; a real browser returns an empty page).

A storage array with no stated port count would have to be invented, so this stays open. If someone can paste the front-end port count for the requested configuration, I will land it the same day.

**Blocked on a port count, not on research effort.** Huawei's public page for the OceanProtect X8100/X9100 gives the port *types* - front-end **10/25/40/100GE and 16/32/64 Gbit/s Fibre Channel**, with 2 or 4 controllers per node and up to 8 in a system - but **not how many front-end ports a node has**. That number depends on the interface-card configuration, and the configuration-specific tables live on `support.huawei.com`, which is behind a login wall I cannot pass (plain fetch returns HTTP 422; a real browser returns an empty page). A storage array with no stated port count would have to be invented, so this stays open. If someone can paste the front-end port count for the requested configuration, I will land it the same day.
Author
Owner

Closing status: blocked on a stated configuration.

Huawei publishes OceanProtect front-end port types - "16/32/64 Gbit/s FC; 10/25/40/100GE" - and a controller count, but not a port count, because how many front-end ports a backup array has is a function of the controllers and interface cards ordered. E8000's page publishes no interface table at all.

The X-series comparison table is otherwise complete (capacity, bandwidth, controllers, cache), which is why this is a narrow blocker rather than an unreadable page.

What would close it: a controller/interface-card configuration, or an authoritative Huawei ordering guide that pins ports per base model. docs/DEVICE_REGISTRY.md §"What is left open, and why" records the same reason for all three (#741, #742, #743).

Closing status: **blocked on a stated configuration.** Huawei publishes OceanProtect front-end port **types** - "16/32/64 Gbit/s FC; 10/25/40/100GE" - and a controller count, but not a port count, because how many front-end ports a backup array has is a function of the controllers and interface cards ordered. E8000's page publishes no interface table at all. The X-series comparison table is otherwise complete (capacity, bandwidth, controllers, cache), which is why this is a narrow blocker rather than an unreadable page. **What would close it:** a controller/interface-card configuration, or an authoritative Huawei ordering guide that pins ports per base model. `docs/DEVICE_REGISTRY.md` §"What is left open, and why" records the same reason for all three (#741, #742, #743).
Author
Owner

Re-checked at the source level. Still blocked on the same field, but the reason recorded on 2026-11-09 - "support.huawei.com is behind a login wall" - is wrong, and that mattered because it made the blocker look unreadable rather than merely unpublished.

Sources checked today (2026-11-17)

  • https://e.huawei.com/en/products/storage/oceanprotect/x100-series (and the DE copy) - X8100/X9100 spec table: controllers per node 2/2/4/4, max controllers 4/4/8/8, Front-End Port Types | 16/32/64 Gbit/s Fibre Channel; 10/25/40/100GE. No port-count row.
  • https://e.huawei.com/marketingcloud/pep/asset/20000001/Material/9f5f91eed2a44fef9ffc21f94fe26982/M3T1A590N1101096260646756423/Huawei%20OceanProtect%20X8000%20Backup%20Storage%20-%20Data%20Sheet.pdf and the all-series data sheet - same shape at 8/16/32/64 Gb FC / 1/10/25/40/100 GbE. No counts.
  • https://support.huawei.com/enterprise/en/data-protection/oceanprotect-x8100-pid-268353477 - HTTP 200; it lists the document set that would carry the census: Product Descriptions EDOC1100368229 (X6000/X8000/X8100), EDOC1100368230 (X9000/X9100), EDOC1100486186 (E8000); Hardware Installation Guides EDOC1100305240, EDOC1100305243, EDOC1100486185.
  • Those six IDs all answer HTTP 404 at https://support.huawei.com/enterprise/en/doc/<ID> (redirect to /view/notFound), because the viewer only serves /enterprise/en/doc/<ID>/<8-hex-section>/<slug> and none of this set's section slugs is indexed.
  • The wall is not a wall: a crawler user-agent gets HTTP 200 with the full server-rendered page. Verified on https://support.huawei.com/enterprise/en/doc/EDOC1100318730/ce945100/cabinet-layout-and-connection-planning (OceanProtect X6000, X8000 1.3.0 Installation Guide, 567,916 bytes) and .../EDOC1100318733/4f560405/... (X3000 1.3.0, 517,177 bytes).

Exact missing fields for X8100 and X9100: interface modules per controller and which slots accept front-end modules, plus the onboard front-end ports the node ships with - equivalently one base ordering configuration ("X8100 node, 2 controllers, 2 x 4-port 25GE IOM per controller").

What would unblock it: a stated configuration of that shape, or any indexed section of EDOC1100368229 / EDOC1100368230 - one reachable page converts both models in one PR.

One trap worth recording: https://support.huawei.com/enterprise/en/doc/EDOC1100484689/bb36e2c/connection-planning-between-the-storage-system-and-application-server looks like the answer - Table 6-1 gives module port counts (GE 4, 10GE 4, 40GE 2, 100GE 2, 16/32/64 Gb FC 4) - but its own title is OceanStor Dorado 8000 and Dorado 18000 V700R001 Site Planning Guide. It is not an OceanProtect document and must not be imported here. Same for the "two interface module slots, 2 x 10GE onboard" text in the X3000 guide: that is the X3000, not the X8100.

PR #1764 records this in research/oceanprotect-x-series-port-census.md; it does not close this issue.

**Re-checked at the source level. Still blocked on the same field, but the reason recorded on 2026-11-09 - "support.huawei.com is behind a login wall" - is wrong, and that mattered because it made the blocker look unreadable rather than merely unpublished.** **Sources checked today (2026-11-17)** - `https://e.huawei.com/en/products/storage/oceanprotect/x100-series` (and the DE copy) - X8100/X9100 spec table: controllers per node 2/2/4/4, max controllers 4/4/8/8, `Front-End Port Types | 16/32/64 Gbit/s Fibre Channel; 10/25/40/100GE`. No port-count row. - `https://e.huawei.com/marketingcloud/pep/asset/20000001/Material/9f5f91eed2a44fef9ffc21f94fe26982/M3T1A590N1101096260646756423/Huawei%20OceanProtect%20X8000%20Backup%20Storage%20-%20Data%20Sheet.pdf` and the all-series data sheet - same shape at 8/16/32/64 Gb FC / 1/10/25/40/100 GbE. No counts. - `https://support.huawei.com/enterprise/en/data-protection/oceanprotect-x8100-pid-268353477` - HTTP 200; it lists the document set that would carry the census: Product Descriptions `EDOC1100368229` (X6000/X8000/X8100), `EDOC1100368230` (X9000/X9100), `EDOC1100486186` (E8000); Hardware Installation Guides `EDOC1100305240`, `EDOC1100305243`, `EDOC1100486185`. - Those six IDs all answer HTTP 404 at `https://support.huawei.com/enterprise/en/doc/<ID>` (redirect to `/view/notFound`), because the viewer only serves `/enterprise/en/doc/<ID>/<8-hex-section>/<slug>` and none of this set's section slugs is indexed. - The wall is not a wall: a crawler user-agent gets HTTP 200 with the full server-rendered page. Verified on `https://support.huawei.com/enterprise/en/doc/EDOC1100318730/ce945100/cabinet-layout-and-connection-planning` (OceanProtect X6000, X8000 1.3.0 Installation Guide, 567,916 bytes) and `.../EDOC1100318733/4f560405/...` (X3000 1.3.0, 517,177 bytes). **Exact missing fields for X8100 and X9100:** interface modules per controller and which slots accept front-end modules, plus the onboard front-end ports the node ships with - equivalently one base ordering configuration ("X8100 node, 2 controllers, 2 x 4-port 25GE IOM per controller"). **What would unblock it:** a stated configuration of that shape, or any indexed section of `EDOC1100368229` / `EDOC1100368230` - one reachable page converts both models in one PR. **One trap worth recording:** `https://support.huawei.com/enterprise/en/doc/EDOC1100484689/bb36e2c/connection-planning-between-the-storage-system-and-application-server` looks like the answer - Table 6-1 gives module port counts (GE 4, 10GE 4, 40GE 2, 100GE 2, 16/32/64 Gb FC 4) - but its own title is *OceanStor Dorado 8000 and Dorado 18000 V700R001 Site Planning Guide*. It is not an OceanProtect document and must not be imported here. Same for the "two interface module slots, 2 x 10GE onboard" text in the X3000 guide: that is the X3000, not the X8100. PR #1764 records this in `research/oceanprotect-x-series-port-census.md`; it does not close this issue.
Author
Owner

The four books that would close this are now identified by ID, and the pair turns out to be two different boxes. PR #1811 adds it to research/oceanprotect-x-series-port-census.md; the issue stays open.

https://support.huawei.com/enterprise/en/data-protection/oceanprotect-x8100-pid-268353477 (200, 420,838 B) splits the request across four documents:

book ID covers
Product Description EDOC1100368229 X6000, X8000, X8100
Product Description EDOC1100368230 X9000, X9100
Hardware Installation Guide EDOC1100305240 X6000, X8000, X8100
Hardware Installation Guide EDOC1100305243 X9000, X9100

Two corrections to what this thread assumes:

  1. The X8100/X9100 data sheet exists and I read it end to end — it is a PowerPoint. Huawei OceanProtect X8100 X9100 Backup Storage Data Sheet V200R001C30.pptx, 5,265,602 B, 5 slides. Slide 4 has one Front-End Port Types row and nothing else: zero occurrences of slot, IOM, interface module or port count across the deck. Huawei storage data sheets are pptx files (the E8000's is too), which is why three passes of PDF-hunting found nothing.
  2. X8100 and X9100 are different enclosures, so this cannot be one spec even when the census arrives. X8100 node: 2 U, 86.1 x 447 x 920 mm, <=45 kg, 2 controllers/node. X9100 node: 4 U, 175 x 447 x 865 mm, <=89 kg, 4 controllers/node. Whichever census is found gets its own record.

Access state of all four books (measured today): doc roots 302 to /view/notFound on support.huawei.com and support.huawei.cn; hedex/hdx.do?docid=<ID>&lang=en returns a 3,728-byte viewer shell; hedex/api/navi/<ID>/first-level-tree answers 401 {"status":401,"message":"NOT LOGIN!"}; a section hash borrowed from the 1.3.0 book (5ee6c5f0/overview) does not transfer. The 1.3.0 book still parses (200, 501,601 B) — that is why X8000 is modelled and X8100/X9100 are not.

Exact missing fields: per model, the controller-enclosure interface-module slot count, which slots accept front-end modules, and the per-module port count for the 10/25/40/100GE and 16/32/64 Gb FC modules Huawei sells for it.

What unblocks it: any one live section of those four EDOCs — one slug opens a whole book, exactly as ce945100 did for X6000/X8000 — or one stated configuration ("X8100 node, 2 controllers, 4-port 25GE module in IOM 0 and IOM 1 of each controller").

**The four books that would close this are now identified by ID, and the pair turns out to be two different boxes.** PR #1811 adds it to `research/oceanprotect-x-series-port-census.md`; the issue stays open. `https://support.huawei.com/enterprise/en/data-protection/oceanprotect-x8100-pid-268353477` (200, 420,838 B) splits the request across four documents: | book | ID | covers | |---|---|---| | Product Description | `EDOC1100368229` | X6000, X8000, **X8100** | | Product Description | `EDOC1100368230` | X9000, **X9100** | | Hardware Installation Guide | `EDOC1100305240` | X6000, X8000, **X8100** | | Hardware Installation Guide | `EDOC1100305243` | X9000, **X9100** | Two corrections to what this thread assumes: 1. **The X8100/X9100 data sheet exists and I read it end to end — it is a PowerPoint.** `Huawei OceanProtect X8100 X9100 Backup Storage Data Sheet V200R001C30.pptx`, 5,265,602 B, 5 slides. Slide 4 has one `Front-End Port Types` row and nothing else: **zero** occurrences of `slot`, `IOM`, `interface module` or `port count` across the deck. Huawei storage data sheets are pptx files (the E8000's is too), which is why three passes of PDF-hunting found nothing. 2. **X8100 and X9100 are different enclosures, so this cannot be one spec even when the census arrives.** X8100 node: 2 U, 86.1 x 447 x 920 mm, <=45 kg, 2 controllers/node. X9100 node: 4 U, 175 x 447 x 865 mm, <=89 kg, 4 controllers/node. Whichever census is found gets its own record. **Access state of all four books (measured today):** doc roots 302 to `/view/notFound` on `support.huawei.com` and `support.huawei.cn`; `hedex/hdx.do?docid=<ID>&lang=en` returns a 3,728-byte viewer shell; `hedex/api/navi/<ID>/first-level-tree` answers `401 {"status":401,"message":"NOT LOGIN!"}`; a section hash borrowed from the 1.3.0 book (5ee6c5f0/overview) does not transfer. The 1.3.0 book still parses (200, 501,601 B) — that is why X8000 is modelled and X8100/X9100 are not. **Exact missing fields:** per model, the controller-enclosure interface-module slot count, which slots accept front-end modules, and the per-module port count for the 10/25/40/100GE and 16/32/64 Gb FC modules Huawei sells for it. **What unblocks it:** any one live section of those four EDOCs — one slug opens a whole book, exactly as `ce945100` did for X6000/X8000 — or one stated configuration ("X8100 node, 2 controllers, 4-port 25GE module in IOM 0 and IOM 1 of each controller").
Author
Owner

Still blocked. This pass found a third anonymous documentation shape, proved it does not belong to OceanProtect, and sourced the X8100/X9100 node table and slot count. PR #1830 records it; issue stays open.

The route (new)
https://info.support.huawei.com/hedex/api/pages/<EDOC>/<viewer>/<issue>/resources/<topic>.html - static HTML, one file per Product Description topic, no login, indexed by search engines. It gives exactly the sentences wanted: "A 100GE interface module provides two 100 Gbit/s optical ports" and "A GE electrical interface module provides four 1 Gbit/s electrical ports".

Why it cannot be used. Those topics are EDOC1100233761 = the OceanStor 5x10/6x10 Product Description (ENISA EUCC-3096-2026-003 cites hedex/hdx.do?docid=EDOC1100233761&id=oceanstor_5x10_031; the topics carry HyperMetro, SmartMobility, Dorado content). prod_desc_mid_* is the OceanStor mid-range family. Importing a 2-port 100GE or 4-port GE module from there into an OceanProtect node is the family-sibling error this catalogue has corrected before - and this source is more dangerous than the Dorado Table 6-1 already noted, because the sentence never names a product.

Access mechanics, so nobody re-walks it: plain curl + Googlebot UA => 403 Access Denied on support.huawei.com/hedex/..., silent rewrite to support.huawei.com/supportindex/ on info.support.huawei.com/hedex/...; r.jina.ai is rewritten the same way; a non-existent topic also lands on the support index, so "200 + index chrome" means no such topic.

Newly sourced from vendor text for X8100/X9100

  • https://e.huawei.com/en/products/storage/oceanprotect/x100-series (200): controllers 2 / 4, CPU 2x2x48 or 64 cores, memory 1 TB / 2 TB, bandwidths, NVMe disk types, front-end port types, C13 vs C19 feed, 86.1x447x920 mm (X8100) / 175x447x865 mm (X9100), <=45 kg / <=89 kg. No port count in any row.
  • Huawei OceanProtect X Series Backup Appliance Technical White Paper, Issue 01 (2026-09-29), 16,574,915 B, material page e.huawei.com/en/documents/products/storage/eded1b265ecc4dfba2fbf8669b0b8213: "Each controller enclosure supports up to 12 hot-swappable interface modules"; "Controllers A and B separately provide six interface module slots"; each X8100 controller has only "two GE management/maintenance ports and one serial port", so every front-end port comes from an IOM. IOM kinds listed (GE electrical, 10GE electrical, 40GE optical, 100GE optical, SmartIO 8/16/32 Gb FC, 10/25 Gbit/s, 4-port 25 Gbit/s RDMA) with no ports-per-module count.

Exact missing fields: ports per 10/25/40/100GE module and per 16/32/64 Gb FC module, from an OceanProtect book; plus one stated configuration of the shape devices/huawei/oceanprotect-x8000.json uses ("4-port 10GE electrical module in IOM 0 and IOM 1 of each controller").

What would close it: any live section of EDOC1100305240 (X6000/X8000/X8100 Hardware Installation Guide), EDOC1100368229 (their Product Description), EDOC1100368230/EDOC1100305243 (X9000/X9100) or EDOC1100486186 (E8000) - one 8-hex slug opens the whole book, exactly as ce945100 did for X6000/X8000 - or a Huawei ordering guide that pins ports per base model. All five still answer HTTP 404 on .../doc/<EDOC>?idPath=... (452,479-byte /view/notFound), so the product page's own query does not open them either.

Related: #742's E8000 data sheet was read end to end this pass (2,407,538 B pptx, 4 slides) and carries only Front-End Port Types 16/32/64 Gbit/s Fibre Channel; 10/25/40/100GE, so that request is unchanged.

**Still blocked. This pass found a third anonymous documentation shape, proved it does not belong to OceanProtect, and sourced the X8100/X9100 node table and slot count. PR #1830 records it; issue stays open.** **The route (new)** `https://info.support.huawei.com/hedex/api/pages/<EDOC>/<viewer>/<issue>/resources/<topic>.html` - static HTML, one file per Product Description topic, no login, indexed by search engines. It gives exactly the sentences wanted: "A 100GE interface module provides two 100 Gbit/s optical ports" and "A GE electrical interface module provides four 1 Gbit/s electrical ports". **Why it cannot be used.** Those topics are `EDOC1100233761` = the **OceanStor 5x10/6x10** Product Description (ENISA EUCC-3096-2026-003 cites `hedex/hdx.do?docid=EDOC1100233761&id=oceanstor_5x10_031`; the topics carry HyperMetro, SmartMobility, Dorado content). `prod_desc_mid_*` is the OceanStor mid-range family. Importing a 2-port 100GE or 4-port GE module from there into an OceanProtect node is the family-sibling error this catalogue has corrected before - and this source is more dangerous than the Dorado Table 6-1 already noted, because the sentence never names a product. Access mechanics, so nobody re-walks it: plain `curl` + Googlebot UA => **403 Access Denied** on `support.huawei.com/hedex/...`, silent rewrite to `support.huawei.com/supportindex/` on `info.support.huawei.com/hedex/...`; `r.jina.ai` is rewritten the same way; a non-existent topic also lands on the support index, so "200 + index chrome" means no such topic. **Newly sourced from vendor text for X8100/X9100** - `https://e.huawei.com/en/products/storage/oceanprotect/x100-series` (200): controllers 2 / 4, CPU 2x2x48 or 64 cores, memory 1 TB / 2 TB, bandwidths, NVMe disk types, front-end port *types*, C13 vs C19 feed, 86.1x447x920 mm (X8100) / 175x447x865 mm (X9100), <=45 kg / <=89 kg. No port count in any row. - `Huawei OceanProtect X Series Backup Appliance Technical White Paper`, Issue 01 (2026-09-29), 16,574,915 B, material page `e.huawei.com/en/documents/products/storage/eded1b265ecc4dfba2fbf8669b0b8213`: "Each controller enclosure supports up to 12 hot-swappable interface modules"; "Controllers A and B separately provide six interface module slots"; each X8100 controller has only "two GE management/maintenance ports and one serial port", so every front-end port comes from an IOM. IOM kinds listed (GE electrical, 10GE electrical, 40GE optical, 100GE optical, SmartIO 8/16/32 Gb FC, 10/25 Gbit/s, 4-port 25 Gbit/s RDMA) with no ports-per-module count. **Exact missing fields:** ports per 10/25/40/100GE module and per 16/32/64 Gb FC module, from an **OceanProtect** book; plus one stated configuration of the shape `devices/huawei/oceanprotect-x8000.json` uses ("4-port 10GE electrical module in IOM 0 and IOM 1 of each controller"). **What would close it:** any live section of `EDOC1100305240` (X6000/X8000/X8100 Hardware Installation Guide), `EDOC1100368229` (their Product Description), `EDOC1100368230`/`EDOC1100305243` (X9000/X9100) or `EDOC1100486186` (E8000) - one 8-hex slug opens the whole book, exactly as `ce945100` did for X6000/X8000 - or a Huawei ordering guide that pins ports per base model. All five still answer HTTP 404 on `.../doc/<EDOC>?idPath=...` (452,479-byte `/view/notFound`), so the product page's own query does not open them either. Related: #742's E8000 data sheet was read end to end this pass (2,407,538 B pptx, 4 slides) and carries only `Front-End Port Types 16/32/64 Gbit/s Fibre Channel; 10/25/40/100GE`, so that request is unchanged.
Author
Owner

Still blocked. This pass read the newest datasheet in full, identified the exact host that serves the census, and renamed the blocker: it is an ALB header rule on info.support.huawei.com, not a login wall. PR #1838 records it in research/oceanprotect-x-series-port-census.md §11; the issue stays open.

Sources read today (vendor only)

  • https://e.huawei.com/en/documents/products/storage/7450a63331f44a35a6d85d42bc2b79ed (200) -> Huawei OceanProtect X8100 X9100 Appliance Data Sheet V200R001C33 (2026 H2), asset https://e-file.huawei.com/marketingcloud/pep/asset/20000001/Material/7450a63331f44a35a6d85d42bc2b79ed/M3T1A590N1319414983213645852/...pdf, 743,373 B, PDF 1.6, 11 pages. Read end to end: the only port row is Front-End Port Types | 16/32/64 Gbit/s Fibre Channel; 10/25/40/100GE. The string slot occurs 0 times and IOM 0 times in the file. (The C30 revision read earlier was a 5-slide pptx; C33 is a real PDF and still answers nothing.)
  • https://support.huawei.com/enterprise/en/data-protection/oceanprotect-x8100-pid-268353477 (200, 420,418 B, crawler UA) - the product page's own document list.

Two things this pass settles

  1. The book ids in this thread are right; the set was re-issued, not replaced. That page maps the current Product Description to EDOC1100368229 and the Hardware Installation Guide to EDOC1100305240; EDOC1100305241's About This Document table reads OceanProtect X3000 | 1.3.0 1.5.0 1.6.0 1.7.0 V200R001C10. So the 1.3.0-era ids now carry "1.x and V200R001" titles over the same hardware - the X8100's IOM rules sit in EDOC1100305240, and the one missing item is a section slug, not a document.
  2. The block is a network rule, and it is measurable. Every info.support.huawei.com/enterprise/en/doc/... URL of this generation - including the section URLs the search index already holds - answers HTTP 302, 355 bytes, Location: https://support.huawei.com/, with header X-ALB-HIT: req-hd-bl, under Chrome 131, Googlebot, Bingbot and curl/8.4.0 user agents, with and without Referer / Accept / Accept-Language / Sec-Fetch-*. r.jina.ai gets the same bounce. The same books on support.huawei.com answer 403 (browser UA, 442-459 B) or 404 /view/notFound (crawler UA, 452,479 B), with or without the page's own ?idPath=7919749|251366268|261607408|251366267|268353477. hedex/hdx.do?docid= is still the identical 3,728-byte SPA shell for EDOC1100368229, EDOC1100305240 and EDOC1100368651, so no viewer/issue pair is recoverable there either. Earlier notes called this a login wall and then a 404; neither is correct, and that mis-framing is why it kept looking re-triable.

Exact missing fields (unchanged): ports per 10/25/40/100GE module and per 16/32/64 Gbit/s FC module from an OceanProtect book, plus the IOM slots that accept front-end modules on X8100 (2 U, 6 IOM per controller) and X9100 (4 U). What closes it: one live section of EDOC1100305240, EDOC1100368229 or EDOC1100368230 fetched from a network the ALB does not bounce - one 8-hex slug opens the whole book, exactly as ce945100 did for X6000/X8000 - or a stated configuration of the shape X8100 node, 2 controllers, 4-port 10GE electrical module in IOM 0 and IOM 1.

Recorded as a search shape, not as evidence: EDOC1100337685 (OceanProtect X9000 1.5.0 Product Description), section 8ed1346b/overview, gives a 4 U slot topology - SMM 0/SMM 1, IOM H0-H13 and L0-L13, back-end only in H5/L5-H8/L8. It names X9000, so it is not imported for X9100.

**Still blocked. This pass read the newest datasheet in full, identified the exact host that serves the census, and renamed the blocker: it is an ALB header rule on `info.support.huawei.com`, not a login wall.** PR #1838 records it in `research/oceanprotect-x-series-port-census.md` §11; the issue stays open. **Sources read today (vendor only)** - `https://e.huawei.com/en/documents/products/storage/7450a63331f44a35a6d85d42bc2b79ed` (200) -> *Huawei OceanProtect X8100 X9100 Appliance Data Sheet **V200R001C33** (2026 H2)*, asset `https://e-file.huawei.com/marketingcloud/pep/asset/20000001/Material/7450a63331f44a35a6d85d42bc2b79ed/M3T1A590N1319414983213645852/...pdf`, 743,373 B, PDF 1.6, **11 pages**. Read end to end: the only port row is `Front-End Port Types | 16/32/64 Gbit/s Fibre Channel; 10/25/40/100GE`. The string `slot` occurs **0** times and `IOM` **0** times in the file. (The C30 revision read earlier was a 5-slide pptx; C33 is a real PDF and still answers nothing.) - `https://support.huawei.com/enterprise/en/data-protection/oceanprotect-x8100-pid-268353477` (200, 420,418 B, crawler UA) - the product page's own document list. **Two things this pass settles** 1. **The book ids in this thread are right; the set was re-issued, not replaced.** That page maps the current Product Description to `EDOC1100368229` and the Hardware Installation Guide to `EDOC1100305240`; `EDOC1100305241`'s About This Document table reads `OceanProtect X3000 | 1.3.0 1.5.0 1.6.0 1.7.0 V200R001C10`. So the 1.3.0-era ids now carry "1.x and V200R001" titles over the same hardware - the X8100's IOM rules sit in `EDOC1100305240`, and the one missing item is a **section slug**, not a document. 2. **The block is a network rule, and it is measurable.** Every `info.support.huawei.com/enterprise/en/doc/...` URL of this generation - including the section URLs the search index already holds - answers **HTTP 302, 355 bytes, `Location: https://support.huawei.com/`**, with header `X-ALB-HIT: req-hd-bl`, under Chrome 131, Googlebot, Bingbot and `curl/8.4.0` user agents, with and without `Referer` / `Accept` / `Accept-Language` / `Sec-Fetch-*`. `r.jina.ai` gets the same bounce. The same books on `support.huawei.com` answer **403** (browser UA, 442-459 B) or **404 `/view/notFound`** (crawler UA, 452,479 B), with or without the page's own `?idPath=7919749|251366268|261607408|251366267|268353477`. `hedex/hdx.do?docid=` is still the identical 3,728-byte SPA shell for `EDOC1100368229`, `EDOC1100305240` and `EDOC1100368651`, so no viewer/issue pair is recoverable there either. Earlier notes called this a login wall and then a 404; neither is correct, and that mis-framing is why it kept looking re-triable. **Exact missing fields (unchanged):** ports per 10/25/40/100GE module and per 16/32/64 Gbit/s FC module from an OceanProtect book, plus the IOM slots that accept front-end modules on **X8100** (2 U, 6 IOM per controller) and **X9100** (4 U). **What closes it:** one live section of `EDOC1100305240`, `EDOC1100368229` or `EDOC1100368230` fetched from a network the ALB does not bounce - one 8-hex slug opens the whole book, exactly as `ce945100` did for X6000/X8000 - or a stated configuration of the shape `X8100 node, 2 controllers, 4-port 10GE electrical module in IOM 0 and IOM 1`. **Recorded as a search shape, not as evidence:** `EDOC1100337685` (*OceanProtect X9000 1.5.0 Product Description*), section `8ed1346b/overview`, gives a 4 U slot topology - `SMM 0`/`SMM 1`, `IOM H0`-`H13` and `L0`-`L13`, back-end only in `H5/L5`-`H8/L8`. It names X9000, so it is **not** imported for X9100.
Author
Owner

Still blocked, and this pass killed the last route that looked like a route: it is not the network. PR #1859 records it in research/oceanprotect-x-series-port-census.md §12; the issue stays open.

§11(c) of that file says the doc host "refuses this network". That is wrong in the way that matters, and it was making the blocker look like something a proxy would fix. One crawler user-agent, one session, minutes apart, same host:

URL result
EDOC1100318730/ce945100/cabinet-layout-and-connection-planning 200, 567,916 B
EDOC1100319420/5ee6c5f0/overview 200, 501,601 B
EDOC1100319420/a9dbbf6/interface-module 200, 485,245 B
EDOC1100305240/216dcdb5/... 302 -> /view/notFound
EDOC1100305242/4dbe1160/... 302 -> /view/notFound
EDOC1100305200/4bd9b2a1/... 302 -> /view/notFound
EDOC1100368229?idPath=… (X6000, X8000, X8100 Product Description) 302 -> /view/notFound
EDOC1100368230?idPath=… (X9000, X9100 Product Description) 302 -> /view/notFound
EDOC1100305237?idPath=… (X6000, X8000, X8100 Quick Installation Guide) 302 -> /view/notFound

So the viewer renders the 1.3.0 "(with Data Backup Feature)" generation to an anonymous crawler and refuses everything outside it. The discriminator is the book, not the host and not our address - which means no ALB bypass, proxy or alternate network will produce EDOC1100368229, EDOC1100368230 or EDOC1100305240; there is no renderable version of them to bypass to.

Three things now closed rather than assumed:

  1. The readable generation cannot be borrowed. EDOC1100319420/a9dbbf6/interface-module re-read in full this pass: X8100 occurs 0 times, X9100 0 times in 485,245 bytes. Its Interface Module chapter links exactly four modules (10GE Electrical, SmartIO, 25 Gbit/s RoCE, 25 Gbit/s RDMA) - the X8000 census already modelled, nothing about X8100's IOM population.
  2. The archive is not a fallback. Wayback CDX holds exactly one capture across the three ids - support.huawei.com/enterprise/en/doc/EDOC1100305240/216dcdb5/cabinet-layout-and-connection-planning, 20250906080606, status 200 - and that capture is a 6,698-byte SPA shell with an empty <title> and no server-rendered text. Its hash 302s live today, so a section hash taken from an index or an archive is revision-scoped and is not a durable pointer to a readable page.
  3. info.support.huawei.com is not a second door. Same path shape, and every variant answers 302, 355 bytes, Location: https://support.huawei.com/ under Chrome 131, Bingbot, Googlebot and curl/8.4.0 - including through r.jina.ai, whose Final URL is https://support.huawei.com/supportindex/.

Exact fields still missing for this issue: the IOM slots that accept front-end interface modules on X8100 (2 U, 2 controllers) and X9100 (4 U, 4 controllers); and the port count of each 10/25/40/100GE and 16/32/64 Gbit/s FC module as fitted to those two models.

What unblocks it, concretely: (a) a reader with a support session pasting any one section of EDOC1100368229, EDOC1100368230 or EDOC1100305240; (b) Huawei's Spare Parts Query output for the model - the Product Description defers the per-model module list to it by name, and it is the one vendor artefact that would pin both fields at once; or (c) a stated configuration of the shape already used for the X8000 spec, X8100 node, 2 controllers, 4-port 10GE electrical module in IOM 0 and IOM 1, which converts in one step using the slot rules transcribed in §5 of the research file.

Not closing, and not guessing: a 2 U frame with an unpublished IOM population would be an empty array, not a census.

**Still blocked, and this pass killed the last route that looked like a route: it is not the network.** PR #1859 records it in `research/oceanprotect-x-series-port-census.md` §12; the issue stays open. `§11(c)` of that file says the doc host "refuses this network". That is wrong in the way that matters, and it was making the blocker look like something a proxy would fix. One crawler user-agent, one session, minutes apart, same host: | URL | result | |---|---| | `EDOC1100318730/ce945100/cabinet-layout-and-connection-planning` | **200, 567,916 B** | | `EDOC1100319420/5ee6c5f0/overview` | **200, 501,601 B** | | `EDOC1100319420/a9dbbf6/interface-module` | **200, 485,245 B** | | `EDOC1100305240/216dcdb5/...` | 302 -> `/view/notFound` | | `EDOC1100305242/4dbe1160/...` | 302 -> `/view/notFound` | | `EDOC1100305200/4bd9b2a1/...` | 302 -> `/view/notFound` | | `EDOC1100368229?idPath=…` (X6000, X8000, X8100 Product Description) | 302 -> `/view/notFound` | | `EDOC1100368230?idPath=…` (X9000, X9100 Product Description) | 302 -> `/view/notFound` | | `EDOC1100305237?idPath=…` (X6000, X8000, X8100 Quick Installation Guide) | 302 -> `/view/notFound` | So the viewer renders the **1.3.0 "(with Data Backup Feature)" generation** to an anonymous crawler and refuses everything outside it. The discriminator is the book, not the host and not our address - which means no ALB bypass, proxy or alternate network will produce `EDOC1100368229`, `EDOC1100368230` or `EDOC1100305240`; there is no renderable version of them to bypass to. **Three things now closed rather than assumed:** 1. **The readable generation cannot be borrowed.** `EDOC1100319420/a9dbbf6/interface-module` re-read in full this pass: `X8100` occurs **0** times, `X9100` **0** times in 485,245 bytes. Its Interface Module chapter links exactly four modules (`10GE Electrical`, `SmartIO`, `25 Gbit/s RoCE`, `25 Gbit/s RDMA`) - the X8000 census already modelled, nothing about X8100's IOM population. 2. **The archive is not a fallback.** Wayback CDX holds exactly one capture across the three ids - `support.huawei.com/enterprise/en/doc/EDOC1100305240/216dcdb5/cabinet-layout-and-connection-planning`, `20250906080606`, status 200 - and that capture is a **6,698-byte SPA shell** with an empty `<title>` and no server-rendered text. Its hash 302s live today, so a section hash taken from an index or an archive is revision-scoped and is not a durable pointer to a readable page. 3. **`info.support.huawei.com` is not a second door.** Same path shape, and every variant answers 302, 355 bytes, `Location: https://support.huawei.com/` under Chrome 131, Bingbot, Googlebot and `curl/8.4.0` - including through `r.jina.ai`, whose `Final URL` is `https://support.huawei.com/supportindex/`. **Exact fields still missing for this issue:** the IOM slots that accept front-end interface modules on **X8100** (2 U, 2 controllers) and **X9100** (4 U, 4 controllers); and the port count of each 10/25/40/100GE and 16/32/64 Gbit/s FC module as fitted to those two models. **What unblocks it, concretely:** (a) a reader with a support session pasting any one section of `EDOC1100368229`, `EDOC1100368230` or `EDOC1100305240`; (b) Huawei's **Spare Parts Query** output for the model - the Product Description defers the per-model module list to it by name, and it is the one vendor artefact that would pin both fields at once; or (c) a stated configuration of the shape already used for the X8000 spec, `X8100 node, 2 controllers, 4-port 10GE electrical module in IOM 0 and IOM 1`, which converts in one step using the slot rules transcribed in §5 of the research file. Not closing, and not guessing: a 2 U frame with an unpublished IOM population would be an empty array, not a census.
Author
Owner

Still blocked on the IOM census. This pass fixed the probe method, re-proved the books absent under it, and found the artefact the vendor itself defers the answer to. PR #1867 records it; the issue stays open.

1. The method, corrected first, because it changes how the earlier readings should be trusted. The doc viewer serves the same 3,800-byte SPA shell to a default curl user-agent for a section that renders and for one that does not:

URL default curl UA Googlebot UA
EDOC1100318730/ce945100/cabinet-layout-and-connection-planning 200, 3,800 B, no text 200, 567,916 B, renders
EDOC1100318733/4f560405/cabinet-layout-and-connection-planning 200, 3,800 B, no text 200, 517,177 B, renders

So a 200 + shell from a non-crawler client means not measured, not reachable. Only a 404 with the /view/notFound body discriminates.

2. Under the UA that renders both books above, in the same session, the X8100/X9100 books are absent - not bounced:

URL result
.../EDOC1100368229 (bare root) 404, 452,479 B
.../EDOC1100368229/interface-module (slug, no hash) 404, 383,570 B
.../EDOC1100368229/a9dbbf6/interface-module (borrowed X8000 hash) 404, 452,479 B
.../EDOC1100368230/a9dbbf6/interface-module 404, 452,479 B
.../EDOC1100305240/216dcdb5/... (the Wayback hash) 404, 452,479 B

Topic hashes are revision-scoped and no slug-without-hash form resolves, so there is no enumeration entry point. No ALB bypass helps: there is no renderable version to bypass to.

3. The frame half is now vendor-sourced and does not need those books. https://e.huawei.com/en/products/storage/oceanprotect/x100-series (200, server-rendered) publishes, per model: 2 controllers X8100 / 4 X9100, max 4 / 8 per system, front-end types 16/32/64 Gbit/s Fibre Channel; 10/25/40/100GE, node 86.1 x 447 x 920 mm (X8100, 2 U) and 175 x 447 x 865 mm (X9100, 4 U), node weight <= 45 / <= 89 kg. No IOM slot count, no per-module port count.

4. New route, and it comes from the vendor's own sentence. The Interface Module chapter of the readable generation says: "For details about the types of hot-swappable interface modules supported by each product model, see Spare Parts Query" -> https://info.support.huawei.com/storage/spareparts/#/home. That page is public (200). It is a JavaScript app (main-ZCQDDWQH.js, 2,605,430 B) whose product/module data loads from an endpoint not in the shell bundle, so it needs a browser session, not a fetch. This now sits ahead of re-trying the books: the vendor defers the per-model module list to it by name.

Exact missing fields, unchanged: the IOM slots that accept front-end interface modules on X8100 (2 U, 2 controllers) and X9100 (4 U, 4 controllers), plus the port count of each 10/25/40/100GE and 16/32/64 Gbit/s FC module as fitted to those models.

What unblocks it, in order: (a) Spare Parts Query for the model, in a browser - one screenshot of the X8100 module list closes it; (b) any rendered section of EDOC1100368229, EDOC1100368230 or EDOC1100305240 pasted by a reader with a support session; (c) a stated configuration of the shape X8100 node, 2 controllers, 4-port 10GE electrical module in IOM 0 and IOM 1, which converts in one step using the slot rules already transcribed in research/oceanprotect-x-series-port-census.md section 5.

Not closing, and not guessing: a 2 U frame with an unpublished IOM population would be an empty array, not a census.

**Still blocked on the IOM census. This pass fixed the probe method, re-proved the books absent under it, and found the artefact the vendor itself defers the answer to. PR #1867 records it; the issue stays open.** **1. The method, corrected first, because it changes how the earlier readings should be trusted.** The doc viewer serves the *same 3,800-byte SPA shell* to a default `curl` user-agent for a section that renders and for one that does not: | URL | default curl UA | Googlebot UA | |---|---|---| | `EDOC1100318730/ce945100/cabinet-layout-and-connection-planning` | 200, **3,800 B**, no text | 200, **567,916 B**, renders | | `EDOC1100318733/4f560405/cabinet-layout-and-connection-planning` | 200, **3,800 B**, no text | 200, **517,177 B**, renders | So a `200 + shell` from a non-crawler client means *not measured*, not *reachable*. Only a `404` with the `/view/notFound` body discriminates. **2. Under the UA that renders both books above, in the same session, the X8100/X9100 books are absent - not bounced:** | URL | result | |---|---| | `.../EDOC1100368229` (bare root) | **404**, 452,479 B | | `.../EDOC1100368229/interface-module` (slug, no hash) | **404**, 383,570 B | | `.../EDOC1100368229/a9dbbf6/interface-module` (borrowed X8000 hash) | **404**, 452,479 B | | `.../EDOC1100368230/a9dbbf6/interface-module` | **404**, 452,479 B | | `.../EDOC1100305240/216dcdb5/...` (the Wayback hash) | **404**, 452,479 B | Topic hashes are revision-scoped and no slug-without-hash form resolves, so there is no enumeration entry point. No ALB bypass helps: there is no renderable version to bypass to. **3. The frame half is now vendor-sourced** and does not need those books. `https://e.huawei.com/en/products/storage/oceanprotect/x100-series` (200, server-rendered) publishes, per model: **2 controllers** X8100 / **4** X9100, max 4 / 8 per system, front-end types `16/32/64 Gbit/s Fibre Channel; 10/25/40/100GE`, node `86.1 x 447 x 920 mm` (X8100, 2 U) and `175 x 447 x 865 mm` (X9100, 4 U), node weight <= 45 / <= 89 kg. No IOM slot count, no per-module port count. **4. New route, and it comes from the vendor's own sentence.** The Interface Module chapter of the readable generation says: *"For details about the types of hot-swappable interface modules supported by each product model, see Spare Parts Query"* -> **`https://info.support.huawei.com/storage/spareparts/#/home`**. That page is public (200). It is a JavaScript app (`main-ZCQDDWQH.js`, 2,605,430 B) whose product/module data loads from an endpoint not in the shell bundle, so it needs a browser session, not a fetch. This now sits **ahead** of re-trying the books: the vendor defers the per-model module list to it by name. **Exact missing fields, unchanged:** the IOM slots that accept front-end interface modules on X8100 (2 U, 2 controllers) and X9100 (4 U, 4 controllers), plus the port count of each 10/25/40/100GE and 16/32/64 Gbit/s FC module as fitted to those models. **What unblocks it, in order:** (a) Spare Parts Query for the model, in a browser - one screenshot of the X8100 module list closes it; (b) any rendered section of `EDOC1100368229`, `EDOC1100368230` or `EDOC1100305240` pasted by a reader with a support session; (c) a stated configuration of the shape `X8100 node, 2 controllers, 4-port 10GE electrical module in IOM 0 and IOM 1`, which converts in one step using the slot rules already transcribed in `research/oceanprotect-x-series-port-census.md` section 5. Not closing, and not guessing: a 2 U frame with an unpublished IOM population would be an empty array, not a census.
Author
Owner

Reopened: this close came from a sentence that said the opposite.

Closed at 16:06:14 by the merge of #1867. That PR's body opens with a line disclaiming the close, and the disclaimer is what triggered it: Forgejo reads a close keyword followed by an issue number anywhere in a pull request body, so a denial of the form not + keyword + #N is executed as an instruction to close #N. Nothing was built.

#742 was closed the same way an hour later by a heading in PR #1873. Both are back open.

So that it is visible next time rather than permanent: scripts/check-closed-issues.js now flags any close that contradicts research/open-device-requests.json, and reads the closed list most-recently-updated first, because the tracker's default order is by issue number and this request sat far below the audit's window. closed-audit.yml runs it on every merge. CONTRIBUTING.md now says to write Refs #N and keep the number away from those keywords.

Status unchanged: blocked on the front-end port count per node, per the comment above.

**Reopened: this close came from a sentence that said the opposite.** Closed at 16:06:14 by the merge of #1867. That PR's body opens with a line disclaiming the close, and the disclaimer is what triggered it: Forgejo reads a close keyword followed by an issue number anywhere in a pull request body, so a denial of the form *not* + keyword + `#N` is executed as an instruction to close #N. Nothing was built. #742 was closed the same way an hour later by a heading in PR #1873. Both are back open. So that it is visible next time rather than permanent: `scripts/check-closed-issues.js` now flags any close that contradicts `research/open-device-requests.json`, and reads the closed list most-recently-updated first, because the tracker's default order is by issue number and this request sat far below the audit's window. `closed-audit.yml` runs it on every merge. CONTRIBUTING.md now says to write `Refs #N` and keep the number away from those keywords. Status unchanged: blocked on the front-end port count per node, per the comment above.
Author
Owner

Reopened: closed by a dry run that was not dry.

Both #742 and #743 were closed at 21:54:24-25 UTC by the scripts/close-merged-issues.js under construction during the PR #1906 work — an early parser that matched the negated sentences in this thread ("Refs #742 ... it does not close the issue"). Its own report attributed the drop to PR #1905's housekeeping; #1905 merged four minutes earlier and closed nothing, so that attribution was wrong. The merged script requires a closing keyword to open a line or a sentence and to be un-negated, and it proposes neither issue today.

Nothing was built for #742, and #743's spec is still open in PR #1902 — so the only correct state for both is open.

**Reopened: closed by a dry run that was not dry.** Both #742 and #743 were closed at 21:54:24-25 UTC by the `scripts/close-merged-issues.js` under construction during the PR #1906 work — an early parser that matched the *negated* sentences in this thread ("Refs #742 ... it does **not** close the issue"). Its own report attributed the drop to PR #1905's housekeeping; #1905 merged four minutes earlier and closed nothing, so that attribution was wrong. The merged script requires a closing keyword to open a line or a sentence and to be un-negated, and it proposes neither issue today. Nothing was built for #742, and #743's spec is still **open in PR #1902** — so the only correct state for both is open.
Author
Owner

Reopened: the close arrived before the pull request that answers it.

Closed 2026-11-17 21:54:25 UTC with no comment. PR #1902 (storage/743, the X8100/X9100 management-plane catalogue entry) is still open, so neither spec is on main yet:

git cat-file -e origin/main:devices/huawei/oceanprotect-x8100.json   -> missing
git cat-file -e origin/main:devices/huawei/oceanprotect-x9100.json   -> missing

The rule the new closer states is that it closes what the catalogue demonstrably answers. At the merge base of this run the catalogue answers neither, so the close was early rather than wrong. When #1902 merges, its closing reference lands this issue closed with a comment naming the file, which is the right order of events.

Status, therefore, is what the comment above says: the management plane is modelled by #1902, and the front-end census stays unmodelled because no reachable vendor book gives a ports-per-interface-module count. If #1902 is abandoned rather than merged, this issue is the only record of the request.

**Reopened: the close arrived before the pull request that answers it.** Closed 2026-11-17 21:54:25 UTC with no comment. PR #1902 (`storage/743`, the X8100/X9100 management-plane catalogue entry) is still **open**, so neither spec is on `main` yet: ``` git cat-file -e origin/main:devices/huawei/oceanprotect-x8100.json -> missing git cat-file -e origin/main:devices/huawei/oceanprotect-x9100.json -> missing ``` The rule the new closer states is that it closes what the catalogue *demonstrably* answers. At the merge base of this run the catalogue answers neither, so the close was early rather than wrong. When #1902 merges, its closing reference lands this issue closed with a comment naming the file, which is the right order of events. Status, therefore, is what the comment above says: the management plane is modelled by #1902, and the front-end census stays unmodelled because no reachable vendor book gives a ports-per-interface-module count. If #1902 is abandoned rather than merged, this issue is the only record of the request.
Sign in to join this conversation.
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#743
No description provided.