Files
truenas-stacks/home/doris-dashboard/docs/network-cutover-master-operator-sheet.md

12 KiB
Raw Blame History

Network Cutover Master Operator Sheet

For Doris: this is the one-page live operator sheet for tomorrow. Use this instead of bouncing between five docs while tired.

Goal: execute the UniFi network redesign with minimal improvisation, explicit stop/go gates, per-port actions, per-device dispositions, and fast rollback discipline.

Architecture:

  • Keep the target lanes clear: Management, Trusted, Servers, IoT, Guest, Cameras.
  • Keep Legacy CIA alive only as a quarantine/sunset lane.
  • Move true service hosts out of Trusted while keeping human endpoints where they belong.
  • Move easy named IoT wins now; quarantine ambiguous junk; Google is in scope, but only through narrow, reversible tests.

Primary references if deeper detail is needed:

  • network-redesign-implementation-runbook.md
  • network-firewall-rule-order.md
  • legacy-cia-triage-worksheet.md
  • unifi-readonly-recon-2026-05-22.md
  • usw-pro-hd-24-cutover-port-sheet.md
  • old-iot-tomorrow-disposition-list.md

1. Live State Snapshot Already Confirmed

Networks currently present:

  • Management -> 10.5.0.1/24
  • Trusted -> VLAN 51 -> 10.5.1.1/24
  • Old IoT -> VLAN 2 -> 192.168.1.1/24
  • IoT -> VLAN 510 -> 10.5.10.1/24
  • Camera -> VLAN 520 -> 10.5.20.1/24
  • Guest -> VLAN 590 -> 10.5.90.1/24

SSIDs currently present:

  • CIA Via -> Old IoT
  • UNEF's Playhouse -> Camera
  • Whiskey Neat Fuck Ice -> Trusted
  • Yer a Wifi Harry -> Trusted

Current client counts:

  • Management: 2
  • Old IoT: 14
  • Camera: 1
  • Trusted: 14

Current hazards:

  • two unnamed espressif devices are on Management via UniFi Wireless on the U7 Pro
  • almost all WiFi clients are on the U7 Pro
  • there are no custom firewall rules/groups yet, so tomorrows segmentation policy is basically greenfield

2. Tomorrows Must-Win Outcomes

By the time we stop, these should be true:

  • Management is materially cleaned up
  • Cameras lane exists and is not polluted by junk placement mistakes
  • Servers lane exists and core hosts are moved or deliberately staged with validation
  • IoT lane exists and easy-value devices are moved
  • Legacy CIA is explicitly a quarantine/sunset lane
  • no broad panic rules were added to compensate for fatigue

In-scope if conditions support it:

  • Google/cast migration work, but only through narrow, reversible tests
  • U6 LR only if needed to stabilize client behavior or complete the design cleanly

Nice-to-have only if smooth:

  • SSID simplification

3. Absolute Rules

  • Keep one stable admin path alive the whole time.
  • Rocinante is the live operator station, so do not move it until the rest of the server batch is done or an alternate admin path is proven.
  • Change one meaningful thing, then validate.
  • Do not bulk-edit live server ports blindly.
  • Nomad, Serenity, and PD can be moved as one controlled batch only because their NFS/shared-service coupling is real; pre-stage everything first, then validate the batch immediately.
  • Do not try to empty Legacy CIA completely if the window gets noisy.
  • Do not move unknown junk into Management, Trusted, or Servers.
  • Do not let Google/cast problems bait you into broad allow rules.
  • If management reachability degrades, rollback first, think second.

4. Minute-by-Minute Execution Spine

T-30 to T-15: prep

  • open this file
  • open UniFi admin
  • open Doris visual artifacts if helpful
  • confirm Rocinante is the live control box
  • confirm fallback route into UniFi in case Rocinante loses the lane mid-change
  • decide whether U6 LR is in scope or explicitly out of scope

Minute 0-10: baseline capture

  • screenshot/export networks
  • screenshot/export SSIDs
  • screenshot/export firewall/rule state
  • screenshot/export port profiles
  • screenshot/export switch port assignments
  • screenshot/export AP mappings
  • snapshot Management client list
  • snapshot Old IoT client list

Stop/go gate:

  • do not proceed until rollback baseline exists

Minute 10-20: create/normalize objects only

  • create/confirm Servers network object
  • create/confirm any missing profiles
  • create/confirm staged SSIDs if needed
  • create/confirm staged firewall objects/rules

Stop/go gate:

  • no clients should have moved yet
  • Trusted admin path still fine

Minute 20-35: management cleanup first

  • identify/address the 2 Management espressif devices
  • ensure infra devices remain on Management intent
  • remove obvious non-infra junk from Management

Stop/go gate:

  • UniFi still sees gateway/switch/APs
  • admin client still reaches controller

Minute 35-50: security cleanup

  • validate Camera lane
  • move Protect chimes if ready
  • confirm doorbell remains healthy

Minute 50-85: core server move wave

  • leave Rocinante in place as the operator station until the main server batch is done
  • keep FlyingDutchman on Trusted unless a later decision explicitly reclassifies it
  • pre-stage the exact three-port sequence: Nomad, Serenity, PD
  • move Nomad, Serenity, and PD in one controlled wave because of NFS/shared-service coupling
  • validate the batch immediately before touching anything else

Minute 85-110: easy IoT wins

  • MyQ
  • LG dryer
  • Samsung FamilyHub
  • Main-Floor ecobee
  • Upstairs ecobee

Minute 110-125: Legacy CIA quarantine pass

  • mark leftovers as quarantine/defer
  • do not force migration of unknowns
  • apply harsh quarantine posture

Minute 125-140: Google test if the window is still stable

  • test one Google/cast device first
  • if ugly, stop and defer

Minute 140-155: guest/SSID cleanup

  • validate Guest
  • disable only stale SSIDs that are truly no longer needed

Minute 155-180: final validation and stop

  • Trusted ok
  • server services ok
  • moved IoT devices ok
  • security ok
  • Management materially cleaner
  • Legacy CIA now explicitly quarantine

5. Management Offender Sheet

These two devices are currently on Management and need identification or removal from that lane:

  1. espressif
  • IP: 10.5.0.123
  • AP: U7 Pro
  • SSID: UniFi Wireless
  • Default call: identify first; if it is smart-junk, move/quarantine off Management
  1. espressif
  • IP: 10.5.0.189
  • AP: U7 Pro
  • SSID: UniFi Wireless
  • Default call: identify first; if it is smart-junk, move/quarantine off Management

Operator note:

  • these are almost certainly exactly the sort of device that should not live on your infrastructure lane

6. Exact Core Port Move Sheet

Leave Alone Unless Positively Necessary

  • USW Pro HD 24 port 1 -> U7 Pro (Upstairs) -> profile Management
  • USW Pro HD 24 port 27 -> likely infra/uplink -> profile Management
  • USW-24-PoE port 24 -> link up on Management, inspect before touching

Port Move Order

  1. USW Pro HD 24 port 2
  • Live host: Rocinante
  • Current profile visibility from API: ambiguous
  • Action: inspect in UI first, then move to Servers if confirmed
  • Validate:
    • correct subnet/gateway
    • reachable from Trusted
    • intended services still work
  1. USW Pro HD 24 port 3
  • Live host: FlyingDutchman
  • Current profile: Trusted
  • Action: keep on Trusted tomorrow unless a later decision explicitly reclassifies it; this is John and Manndra's gaming PC, not a homelab service host by default
  • Validate:
    • host still reachable
    • services behave as expected
  • Safe default: leave it in Trusted
  1. USW Pro HD 24 port 22
  • Live host: Nomad
  • Current profile: Trusted
  • Action: move as part of the controlled three-host server batch
  • Validate:
    • expected address
    • services reachable from Trusted
  1. USW Pro HD 24 port 24
  • Live host: Serenity
  • Current profile: Trusted
  • Action: move as part of the controlled three-host server batch
  • Validate:
    • expected address
    • intended services reachable
  1. USW Pro HD 24 port 23
  • Live host: PlausibleDeniability
  • Current profile: Trusted
  • Action: move as part of the controlled three-host server batch after Nomad and Serenity are ready
  • Validate:
    • expected address
    • SSH/admin path still works
    • critical homelab services still work
  • Warning:
    • because Nomad, Serenity, and PD are coupled by active shares/services, treat them as one planned wave rather than three unrelated experiments
    • Rocinante should stay as the control box until that wave is complete

7. Old IoT Device Disposition Sheet

Migrate Now

  1. MyQ-29B -> 192.168.1.130 -> target IoT
  2. LG_Smart_Dryer2_open -> 192.168.1.186 -> target IoT
  3. Samsung-FamilyHub -> 192.168.1.149 -> target IoT
  4. Main-Floor ecobee -> 192.168.1.102 -> target IoT
  5. Upstairs ecobee -> 192.168.1.131 -> target IoT

Validation for each:

  • app still works
  • device online
  • no broad helper rule was needed

Migrate Later Only If Calm

  1. Google-Home-Mini -> 192.168.1.185
  2. 3c:8d:20:f3:92:36 -> Google device -> 192.168.1.192
  3. 90:ca:fa:b6:7f:6e -> Google device -> 192.168.1.129

Rule:

  • Google is in scope tomorrow, but start with one Google-class device only unless everything is smooth
  • if it demands broad discovery hacks, stop and defer

Quarantine Tomorrow

  1. 5c:61:99:41:73:40 -> unknown Cloud Network device
  2. 60:74:f4:54:fd:ec -> Private/randomized
  3. 60:74:f4:7b:6a:11 -> Private/randomized
  4. c0:f5:35:20:5d:94 -> AMPAK device
  5. d4:ad:fc:60:90:6a -> Intellirocks
  6. d4:ad:fc:ea:7f:65 -> Intellirocks

Rule:

  • these do not earn clean-IoT membership by being merely online
  • leave them on Legacy CIA quarantine until identified better

8. Camera/Security Reality Check

Already true:

  • front-doorbell is already on Camera at 10.5.20.217

Tomorrows security work is therefore mainly:

  • move/clean up chimes
  • tighten policy
  • preserve the doorbells healthy state

9. Firewall Build Priorities

Build these first:

  • established/related allow
  • invalid drop
  • management shield
  • Trusted admin -> Management allow
  • Trusted -> Servers allow
  • Guest -> internet only
  • IoT -> DNS/NTP/internet + specific helpers only
  • Camera -> DNS/NTP/internet/Protect only
  • Legacy CIA -> DNS/NTP/internet only + explicit one-offs only
  • broad internal denies for Guest/IoT/Camera/Legacy CIA

Do not do tomorrow unless proven necessary:

  • broad mDNS/cast trust exceptions
  • loose IoT -> Servers any
  • loose Legacy CIA -> Trusted any

10. Validation Gates

After each meaningful change, check the smallest thing that proves success.

After management changes

  • UniFi admin still reachable
  • gateway/switch/AP still visible
  • no surprise loss of wireless control

After each server port move

  • host gets correct subnet/gateway
  • reachable from Trusted
  • expected service path works

After the Nomad/Serenity/PD batch

  • NFS/shared-service relationships recover cleanly
  • PD remains reachable from Rocinante/admin path
  • no cross-server dependency is hanging half-broken

After each IoT move

  • device rejoins expected SSID/VLAN
  • app/control works
  • no broad workaround rule added

After security changes

  • doorbell/chime online
  • expected app/admin behavior works

After quarantine posture changes

  • Legacy CIA devices still have minimum acceptable functionality
  • they do not have new local trust

11. Fast Rollback Strip

If anything goes sideways:

  1. revert the last moved device/port first
  2. revert the last SSID/VLAN assignment second
  3. revert the last firewall rule/order change third
  4. restore management-plane reachability before doing anything clever
  5. do not stack fresh changes on top of confusion

Immediate rollback triggers:

  • loss of UniFi management-plane access
  • AP/switch disappears unexpectedly
  • moved server loses reachability and root cause is not obvious quickly
  • critical household function breaks and cannot be explained fast
  • Google weirdness starts baiting sloppy panic rules

12. Explicit Stop Condition

Stop when:

  • Management is materially cleaner
  • Cameras lane is sane
  • core server moves are done or deliberately deferred with reasons
  • easy-value IoT moves are done
  • Legacy CIA is explicitly a quarantine/sunset lane
  • the next remaining work item smells like “heroics” instead of “mechanical completion”

That is success. Not every leftover wart has to die tomorrow.