Sync network redesign repo state and cutover artifacts

This commit is contained in:
Fizzlepoof
2026-05-22 20:16:21 +00:00
parent 4b1e3d4061
commit 9e6dde7695
59 changed files with 5389 additions and 244 deletions

View File

@@ -0,0 +1,135 @@
# Server Batch Validation Checklist
Use this immediately after moving Nomad, Serenity, and PD.
Do not start Google nonsense, SSID cleanup, or extra cleanup until this is green.
Primary control path:
- operator box: Rocinante
- critical dependency group: Nomad + Serenity + PlausibleDeniability
---
## Before the batch
- [ ] Rocinante is stable and currently has UniFi/admin access.
- [ ] Fallback admin path exists if Rocinante drops off.
- [ ] Current port/profile screenshots taken for:
- [ ] Port 22 -> Nomad
- [ ] Port 24 -> Serenity
- [ ] Port 23 -> PD
- [ ] Expected target subnet/gateway for Servers VLAN is written down.
- [ ] You are ready to unwind in reverse order if needed.
Move order:
1. Nomad
2. Serenity
3. PD
Rollback order if ugly:
1. PD
2. Serenity
3. Nomad
---
## Per-host quick checks during the wave
### After Nomad move
- [ ] Link came back
- [ ] Nomad got expected Servers IP/subnet/gateway
- [ ] Rocinante can still reach Nomad
- [ ] Expected Nomad-hosted service(s) respond
- [ ] No obvious share/dependency explosion yet
### After Serenity move
- [ ] Link came back
- [ ] Serenity got expected Servers IP/subnet/gateway
- [ ] Rocinante can still reach Serenity
- [ ] Expected Serenity-hosted service(s) respond
- [ ] No obvious share/dependency explosion yet
### After PD move
- [ ] Link came back
- [ ] PD got expected Servers IP/subnet/gateway
- [ ] Rocinante can still reach PD
- [ ] SSH/admin path to PD still works
- [ ] Critical PD-hosted homelab services respond
---
## Immediate batch validation
Do all of these before touching anything else:
### Reachability
- [ ] Rocinante -> Nomad reachable
- [ ] Rocinante -> Serenity reachable
- [ ] Rocinante -> PD reachable
### IP sanity
- [ ] Nomad is on the intended Servers subnet
- [ ] Serenity is on the intended Servers subnet
- [ ] PD is on the intended Servers subnet
- [ ] Default gateway is correct for all three
### Cross-host dependency sanity
- [ ] NFS mounts/shares recovered cleanly
- [ ] No host is hanging on stale routes or dead mount targets
- [ ] Any name-resolution dependency still works as expected
- [ ] No authentication/admin path you need for the rest of the window disappeared
### Service sanity
- [ ] Doris/dashboard path(s) you need are still reachable
- [ ] Any storage-backed services needed for the live window still work
- [ ] Nothing critical is stuck waiting on a dead share
### Control-plane sanity
- [ ] UniFi is still reachable from Rocinante
- [ ] Switch/AP visibility still looks normal
- [ ] No management-plane surprise coincided with the server batch
---
## Stop conditions
Stop and unwind if any of these are true:
- [ ] PD is not reachable from Rocinante
- [ ] NFS/share recovery is unclear after a short wait
- [ ] You cannot tell whether the problem is routing, DNS, or service binding
- [ ] UniFi/admin access is degraded at the same time
- [ ] You are guessing instead of validating
---
## Reverse-order rollback strip
If the batch is bad and the reason is not immediately obvious:
1. Revert PD port/profile first.
- [ ] PD returns to prior lane/profile
- [ ] PD admin path restored
2. Revert Serenity port/profile second.
- [ ] Serenity returns to prior lane/profile
3. Revert Nomad port/profile third.
- [ ] Nomad returns to prior lane/profile
4. Re-check:
- [ ] Rocinante -> PD
- [ ] Rocinante -> Serenity
- [ ] Rocinante -> Nomad
- [ ] UniFi/admin access stable again
Do not continue the broader migration until this set is stable.
---
## When this batch counts as successful
- [ ] Nomad moved successfully
- [ ] Serenity moved successfully
- [ ] PD moved successfully
- [ ] Rocinante still has the admin path you need
- [ ] Shares/services recovered cleanly enough to continue
- [ ] No emergency broad firewall workaround was needed
If all six are true, move on.
If not, stop pretending and fix or roll back first.