# A2: adding corvid-c to the register

**Ticket** PIR-0918 / action A2 · **Owner** platform, on-call rota ·
**Status** open, reopened · **Ref** PIR-0918 (quorum held), fleet-inventory

The action item reads: *Add corvid-c to the asset register.* The blocker is
one line in the register's schema: **serial number required; the register
does not accept empty.** I have the address (192.0.2.73), the cluster it
votes in, the build date. What I do not have is the eleven characters the
form wants. This ticket is about the eleven characters.

## The probes, in order

| probe | expected | result |
| --- | --- | --- |
| `ipmitool -H 192.0.2.73 fru print` | serial | answers in 0.31 ms. Every field present. Every field zero-filled: `Chassis Serial 000000000000000000000000000000` |
| PDU row 9 outlet map | draw | all outlets mapped, none unmapped, none unexplained |
| procurement PO-21-0044 | serials | two servers: SM-2U-441902, SM-2U-441903. Packing list: two cartons. The invoice: two |
| the other two members' serials | - | corvid-a SM-2U-441902, corvid-b SM-2U-441903 |

The BMC answers. It answers cleanly, at the latency the review measured at
zero deviation across five years. It reports a machine with the correct
shape and the serial field holding zeros, which is what a FRU reads like
before the factory writes it, and nothing has ever written it.

## The hinge

The PO numbers are consecutive. SM-2U-44190**2**, SM-2U-44190**3**. The
register's validator pattern is `SM-2U-[0-9]{6}`. At 01:40 on the 12th I had
the form open on my laptop with `SM-2U-441904` in the serial field and the
row otherwise complete, because the row otherwise was true.

Then I ran the validator against the draft, because I always run the
validator. It passed. I ran the downstream checks: the schema accepted the
commit; the warranty table joined the serial and returned coverage dates
from PO-21-0044's fiscal year; the rack diagram tool placed U16 in row 4
because I left the position blank and it fills blanks. The dashboard showed
a third corvid. Nothing anywhere objected. The only thing in this estate
that ever said *no* to corvid-c was m.reyes's 2024 commit message, and the
register had accepted that too.

I closed the laptop without committing. I want it on the record that the
form would have taken it, all the way down, and that the reason the register
is trusted is not the register.

## Status now

Changed to: **blocked - serial does not exist.** `fleet-inventory` opened
PR #124 at 04:00 this morning, same branch name as the other 123, and I am
assigned. Its diff has the same two filled columns and five empty ones,
because the job will not invent what it cannot observe, and now it is not
alone, and I am not sure I would have passed it.

See also [[stories/quorum]], [[stories/inventory-diff]],
[[stories/axfr-from-53]].

Source: the pages linked above, read 2026-09-13. Invented.