RayNu-V — Path to LOI

Tracker · synced from docs/loihda.md

Months to Bar A

1.5 mo

Bar A · was 1.5 · Bar B 3.5 mo

Overall LOI readiness: 38%
Bar A ETA: 2026-11
Bar B ETA: 2026-12
Confidence: medium
Updated: 2026-09-14

Bar A · dedicated-box 42%

Bar B · RAID fleet 18%

Overall LOI readiness 38%

These numbers are a baseline, not a close. They shrink only when COM2 prints a persist marker, TLS is reachable, or a spare PERC virtual disk actually holds a guest. Nested QEMU and host tests do not move the month count.

Definition

Two letters, not one

A Letter of Intent here is a fleet owner saying they intend to run RayNu-V in non-production on named iron. It is not a purchase order, not 1.0, not cluster, not Windows. The two bars are separate conversations; presenting them as one overstates what is ready.

Bar A · 42%

Dedicated box

One PowerEdge we own, or they can afford to dedicate. The guest disk must survive Force Off of the hypervisor. HTTPS so the management plane can be shown to InfoSec. Alpine-on-virtio is an acceptable first guest if the SKU card says so.

Product effect: a lab handshake we can take without promising to replace their RAID array.

Bar B · 18%

RAID fleet

Keep the iron. Replace the licensed hypervisor. Guest and boot disks already live on PERC H740P virtual disks — that is most of the R640 installed base. USB persist does not close this. Formatting the lab Ubuntu array does not close this.

Product effect: this is the “extend the life of the fleet you already paid for” pitch. Until a spare VD holds a guest across Force Off, we do not take this conversation.

The pieces

What has to be true, and why it matters

Everest HDA scored the loop. LOIHDA scores the consequences of each leftover. A feature can be in the binary and still overstate what a buyer can use.

Everest loop 100%

Closed on iron 11 September 2026: SPA Start of RayNu-F → Alpine setup-disk → ISO-INSTALL-OK → F7 → DISK-BOOT-OK → login:. There is a thing a buyer can watch. Without this, an LOI is a research pitch. This piece does not mean disks survive a hypervisor reboot, or that management is encrypted.
Persist across HV reboot 70%

Today the install disk is leftover DRAM above the precise EPT pool. A guest reset keeps it. A hypervisor Force Off zeros it. Nested File persist is closed; USB/NVMe DurableLun I/O is host-ready; iron COM2 is not. The remaining 30% is one Force Off on a ≥16 GiB stick that is not the ESP Cruzer. A buyer who reboots RayNu-V to recover the box must not lose the VM — that is the first thing they will do.
SKU card 25%

One page: what ships, what does not. Dedicated-box versus fleet. USB versus PERC. Alpine-patched versus unmodified. No cluster. No Windows. This page is the start of that card. Until persist and TLS close, the card has to say “not yet.” Overstated copy here does not survive a second diligence meeting.
TLS 8%

Everest closed on plaintext HTTP on a bring-up LAN. That was an honest deferral, not a product. A CISO will ask whether management is encrypted. The answer today is no. A dedicated-box LOI that skips TLS is a handshake in a lab, not something InfoSec countersigns.
Auth beyond bring-up 28%

raynu-v-bringup is a lab latch. ESP auth.token can override it; iron still used the latch. Fine on a box you physically control. Not fine if two people on the LAN can start VMs. Soft for Bar A if named; not silent.
Guest console UI 22%

Serial already typed the Alpine installer. The SPA shows logs. Nobody has a keyboard in the browser yet. Operators should not need iDRAC SOL to log in. Soft for the first dedicated-box LOI. Hard for anyone who thinks this is vSphere.
PERC RAID I/O 15%

The R640 “HDD” is Ubuntu on a PERC H740P. The mapper already classifies RAID and prints skip PERC so we do not destroy the lab OS. Classification is not I/O. There is no MegaRAID mailbox driver. Firmware RAID BlockIo dies at ExitBootServices. This 15% is census plus classification. Bar B waits on a spare virtual disk — not the Ubuntu standing boot.
Unmodified media 20%

Iron Alpine still rides an ISO patcher and serial auto-answer. Fine for a dedicated-box demo if the SKU card says so. A fleet claim of “unmodified Linux ISO” is false until that patcher is gone. Windows is later still, and is not a peer of TLS.

NOW · M8.0-mech

Force Off is the first honest test

Nested persist already rebooted Alpine without setup-disk. Nested QEMU is not a PowerEdge. The next print that moves Bar A is COM2 after a real Force Off, with virtio still on a DurableLun.

need:  ≥16 GiB USB, different port from the LogiLink ESP
refuse: ESP Cruzer (2–8 GiB) · Ubuntu-on-PERC · leftover DRAM
wait:   durable LUN usb|nvme I/O ready
        virtio-blk … (durable LUN …)
then:   F11 this persist EFI only
close:  RAYNU-V-M8-DISK-PERSIST-OK   ← iron COM2 only
never:  host / CI / nested printing that marker

Rollback remains GitHub Latest v0.1.0-everest-closed — do not flash a persist prototype as the known-good. Do not format the Ubuntu PERC.

Scope

What this tracker does not claim

Open work stays named. These boundaries keep a dedicated-box conversation distinct from a fleet replacement, and keep Everest closed while M8 continues.

A USB persist demonstration is the dedicated-box mechanism. RAID-fleet LOI waits on PERC virtual-disk persist.
TLS, unmodified ISO install, and cluster remain open and are not presented as closed.
Everest remains closed. A patched Alpine ISO is a named residual, not a reason to reopen the summit.
Iron install and persist markers are printed on R640 COM2 only. Host and CI do not emit them.
Latitude and nested QEMU are not PowerEdge R640 evidence.

Full living LOIHDA Everest HDA (closed)

Numbers load from site/loi.json, regenerated from docs/loihda.md by ./tools/sync-loihda-site.sh. Last commit: PENDING. Updated 2026-09-14.