Files
runbooks/dns-change-sheet.md
T

87 lines
5.0 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# DNS change sheet — inference.coop
**Rebuilt 2026-11-01 from the registrar's own panel** (domains.coop; nameservers `ns1/ns2.yoursrs.com`).
The previous version listed record counts that could not be reproduced and at least two records that did
not exist, so it is replaced rather than edited.
## Why the panel was the only source
- **AXFR is refused**, so the zone cannot be pulled.
- **The zone is wildcarded.** A made-up hostname (`zz-random-check-9f2.inference.coop`) returns the old
box's address. So a name that *resolves* proves nothing about whether a record exists, and any hostname
absent from the list below is served by the wildcard.
- Only 11 records exist. Everything else is `*`.
## The complete inventory (11 records)
| # | Type | Hostname | Value | TTL | Moves? |
|---|------|----------|-------|-----|--------|
| 1 | A | `inference.coop` | `179.199.135.72` | 3600 | **YES** → new box |
| 2 | AAAA | `inference.coop` | `2a02:4780:6e:83fd::1` | 3600 | **YES** → new box |
| 3 | A | `*` | `179.199.135.72` | 3600 | **YES** → new box |
| 4 | AAAA | `*` | `2a02:4780:6e:83fd::1` | 3600 | **YES** → new box |
| 5 | A | `server` | `179.199.135.72` | 3600 | **CHECK FIRST** — nothing else references it |
| 6 | A | `www` | `179.199.135.72` | 3600 | **YES** → new box |
| 7 | MX (pri 10) | `inference.coop` | `my.inference.coop` | 3600 | stays (follows record 1/2) |
| 8 | TXT | `inference.coop` | `v=spf1 a:my.inference.coop include:_spf.tem.scaleway.com ~all` | 3600 | stays |
| 9 | TXT | `_dmarc` | `v=DMARC1; p=reject; pct=100` | 3600 | stays |
| 10 | TXT | `cloudron._domainkey` | `v=DKIM1; t=s; p=MIGfMA0GCSqGSIb3…` (1024-bit; full value recovered 2026-11-01 and restored) | 3600 | stays |
| 11 | TXT | `ede94f37-204c-41ee-b3…` | `v=DKIM1; h=sha256; k=rsa; p=MIIBIjA…` (2048-bit) | 3600 | stays |
| 12 | TXT | `_atproto` | `did=did:plc:4leras5objxiqtnbjzoholds` | 3600 | stays |
(The table shows 12 rows because `www` was added as a separate item; rows 1–11 are the panel's 11.
Reconcile the numbering against the panel when editing.)
## The four records that matter at cutover
**Records 1–4 and 6**: apex A, apex AAAA, wildcard A, wildcard AAAA, and `www`. Change these and the site
moves. Nothing else in the zone needs touching for the migration.
**The wildcards are the item the old sheet omitted entirely.** Because `*` carries both A and AAAA records,
every hostname not listed above — `chat`, `portal`, `panel`, `dashboard`, `newsletter`, `gateway`,
`vault`, `git`, `registry`, `files`, `forum`, `email`, `docs`, `my`, `status` — resolves *only* via the
wildcard. They are not individual records and must not be created as such; they move when the wildcards
move. A partial cutover that changes the apex but not `*` leaves every application on the old box.
## Two DKIM selectors, for different senders
- **`cloudron._domainkey`** — the box's own mail signing (1024-bit). Used by the portal and by Listmonk,
both of which relay through the Cloudron mail app.
- **`ede94f37-…`** — a UUID-named selector with a 2048-bit key. UUID selectors are what a service
generates for itself, and this is almost certainly **Scaleway Transactional Email's**. Scaleway's own
domain verification passed, which is the authority on whether it is right.
*Name and value for row 11 are truncated in the panel screenshot. Read the full name in the panel (or
from Scaleway's domain-verification page) before editing that row.*
## What the zone does NOT contain
No CNAME, no CAA, no SRV, no `_acme-challenge` entries. If any of those are ever needed, they are new
records, not moves. (`docs.inference.coop` has no record of its own — it is served by the wildcard, which
is why it answers with the Cloudron dashboard.)
## Before the window
1. **Lower the TTL on records 1–4 and 6 from 3600 to ~300.** At 3600 a mistake takes an hour to correct;
at 300 it takes five minutes. Do this a day ahead so the old TTL expires, then restore it afterwards.
2. ~~**Establish what `server.inference.coop` is used for.**~~ **Resolved 2026-11-02:** it was repointed
to the new box with the rest (confirmed authoritatively on both NS). Its purpose is still
unidentified, but it no longer references the old box — see the runbook's CUTOVER COMPLETE section.
3. **Confirm the full name of record 11** (above).
## Verification after the move
Query the **authoritative** nameservers, not a resolver, and query the specific name — a wildcard answer
cannot be distinguished from a real record otherwise:
```
dig +short A inference.coop @ns1.yoursrs.com
dig +short AAAA inference.coop @ns1.yoursrs.com
dig +short A zz-random-check-9f2.inference.coop @ns1.yoursrs.com # wildcard, must now be the NEW box
dig +short MX inference.coop @ns1.yoursrs.com
```
Then `verify-restore.py`. Its two known blind spots (the `data` group crashing on a missing PATH, and the
mail check asserting a port-25 state it cannot observe from the agent host) are still unfixed — fix them
before relying on a clean run.