sync: auto-sync from HOWARD-HOME at 2026-07-10 15:11:39

Author: Howard Enos
Machine: HOWARD-HOME
Timestamp: 2026-07-10 15:11:39
This commit is contained in:
2026-07-10 15:12:08 -07:00
parent 830a86ea95
commit 46d2ae7a5b

View File

@@ -0,0 +1,166 @@
## User
- **User:** Howard Enos (howard)
- **Machine:** Howard-Home
- **Role:** tech
# The Prairie Schooner — Server Migration (TPS-SERVER -> TPS-SVR)
> Work performed 2026-07-07; logged 2026-07-10. Migration is PAUSED mid-way — see Pending.
## Session Summary
Diagnosed the two Prairie Schooner servers to plan replacing the aging **TPS-SERVER**
(Windows Server 2016 Essentials, Lenovo ThinkServer TS140, 3.8 GB RAM, C: 87% full,
backup dead since 2018) with **TPS-SVR** (Windows Server 2019 Standard, HPE ProLiant
MicroServer Gen10 Plus v2, 32 GB RAM, 5.3 TB). TPS-SERVER was the sole DC (all 5 FSMO),
Enterprise Root CA, DNS, file server, and QuickBooks Enterprise 24 database host. TPS-SVR
had roles pre-installed but was never configured (AD DS role present, never promoted;
DHCP/DNS installed but inert). Domain `TPS.local`, single-domain forest, functional level
2016 (schema v87).
Started with data protection. A Windows Server Backup BMR to `\\TPS-SVR\TPSBackup` failed
at ~12% ("media is write protected" — the known WSB-VHDX-over-SMB fragility), so pivoted
to robocopy pre-seed of the file/QuickBooks data (which doubles as migration step 5).
Pre-seeded `C:\Share` (Quickbooks 144 GB + Deployment 22 GB) TPS-SERVER -> `\\TPS-SVR\Share`.
Rebooted the three workstations (Mel-PC, TPS-JAYMI, TPS-Tina) with a 2-minute warning to
release QuickBooks client locks, then stopped the server QB service briefly to copy a clean,
consistent `schoonerQB2025.QBW` (3 GB, verified byte-for-byte), and restarted QB so users
resumed on the old server.
Ran a full role audit of TPS-SERVER: no shared printers, no folder redirection/roaming/home
dirs/logon scripts, no real QuickBooks Web Connector integration (only Intuit test samples),
DFS namespace not published. The Q: drive map (`\\tps-server\quickbooks`) and software-deploy
GPOs are all keyed to the name `tps-server`, and QuickBooks references the host by IP in the
`.ND` (`ServerIp=192.168.1.125`) — but the `.ND` regenerates when the hosting server changes.
Promoted TPS-SVR to a domain controller. `Test-ADDSDomainControllerInstallation` first failed
because the built-in Administrator was disabled, then because its password did not match the
value Howard provided; enabled + reset Administrator (via guru) to `r3tr0gradE99#`, test passed,
ran the real promotion as a detached SYSTEM task. Promotion succeeded; TPS-SVR rebooted and came
up as a healthy 2nd DC + GC + DNS, AD/DNS fully replicated (17 users, 11 computers, 6 GPOs,
zones), replication 0 errors both ways. Set TPS-SVR IPv4 static (192.168.1.135), cleaned its
DNS/forwarders. Re-disabled Administrator.
Hit a blocking constraint: **no SonicWall access.** The SonicWall (192.168.1.1) is gateway +
DHCP and hands clients DNS = 192.168.1.125 (old DC); we cannot change that, so retiring the old
DC would break client DNS unless the new server assumes .125. Howard decided to replace the
SonicWall with a UniFi gateway (to regain DHCP/DNS control) and PAUSE the migration. Left the
OLD server as primary (all FSMO, QB host, file server, client DNS), new server as an idle
healthy replica. Verified stable, cleaned up one-time tasks, saved resume state to the wiki.
## Key Decisions
- **robocopy pre-seed instead of BMR image** — WSB BMR to an SMB share failed ("media write
protected"); robocopy is reliable and doubles as a migration step. Copied data+timestamps
only (`/COPY:DAT`), ACLs deferred to cutover (machine account can't set ownership remotely).
- **Clean QB copy via brief service stop** — rebooted workstations to drop client locks, stopped
`QuickBooksDB34`+`QBCFMonitorService` on the server to prevent re-lock during the ~3 GB copy,
restarted after so users resume on the old server. Final `.QBW` delta re-copied at cutover.
- **Alias approach over rename/re-IP (initially)** — keep TPS-SVR's own name/IP and add
`TPS-SERVER` as a DNS+SPN alias, avoiding a risky live-DC rename and IP collision.
- **REVERSED by SonicWall lockout** — since we can't change the DHCP-handed DNS (.125), the new
server must end up owning .125 so client DNS survives old-DC retirement. Done safely: assign
.125 only AFTER the old DC is demoted+powered off (no collision). Howard's original instinct
(match the old IP) was correct; UniFi swap will restore flexibility.
- **Enable+reset built-in Administrator temporarily** (Howard chose over elevating guru) — the
only account with Enterprise/Schema Admin for the v87->v88 schema adprep. Re-disabled after.
- **Promotion as a detached scheduled task** with `-NoRebootOnCompletion` — a reaper-killed RMM
command mid-dcpromo would leave a half-promoted DC; detached task avoids that and gives a clean
success signal before a controlled reboot.
- **CA = migrate, not decommission** — Enterprise Root CA still auto-enrolls DC certs (fresh
DomainController cert exp 2027). Migrate under the preserved name. MUST test key export first
(key marked non-exportable, on Microsoft Software KSP).
## Problems Encountered
- **WSB BMR failed at ~12% ("media is write protected")** — WSB VHDX-over-SMB fragility.
Resolved by pivoting to robocopy for data; system-state/BMR deferred.
- **False disk-full alarm** — a `Get-Volume` reading showed 55 GB free mid-copy (impossible;
source is 405 GB). Stopped the copy as a precaution; verified no symlink loop, disk actually
4877 GB free — glitched reading. Resumed; copy completed (166.55 GB / 113,971 files, verified).
- **`Test-ADDSDomainControllerInstallation` STATUS: Error, empty `.Messages`** — real error was in
the warning/verbose stream (`*>&1`): first "account is currently disabled" (built-in Admin
disabled), then "user name or password is incorrect" (provided pw wasn't Administrator's).
Enabled + reset Administrator via guru; test then passed.
- **iconv missing on Howard-Home** — `ps-encoded.sh` failed; used `jq -n --arg` and base64 for
script/credential delivery instead.
- **No SonicWall access** — blocks the client-DNS cutover; Howard replacing it with UniFi.
## Configuration Changes
- **TPS-SVR promoted to DC** (`Install-ADDSDomainController`, DomainName TPS.local, InstallDns,
GC, Site Default-First-Site-Name, NTDS/SYSVOL default C:\ paths). Forest schema adprep'd v87->v88.
- **TPS-SVR IPv4 set static:** 192.168.1.135/24, gw 192.168.1.1; NIC DNS = 192.168.1.125,127.0.0.1;
DNS server forwarders = 1.1.1.1,8.8.8.8 (removed dead Norton ConnectSafe 199.85.x).
- **TPS-SVR shares created:** `\\TPS-SVR\Share` (C:\Share, data pre-seed target),
`\\TPS-SVR\TPSBackup` (C:\Backups\TPS-SERVER, failed BMR target). Data staged in C:\Share.
- **Domain built-in Administrator:** enabled -> password reset to provided value -> RE-DISABLED
(net state: disabled, as before, but password now known/vaulted).
- **One-time scheduled tasks removed** post-run: TPS-BMR-Backup, TPS-Preseed-Copy (TPS-SERVER),
TPS-DCPromo (TPS-SVR).
- Wiki: `wiki/clients/the-prairie-schooner.md` — added migration-status + resume checklist.
- No changes to TPS-SERVER roles/FSMO/shares — it remains the untouched primary.
## Credentials & Secrets
All vaulted (SOPS). Domain `TPS.local`:
- `clients/prairie-schooner/tps-domain-admin` — user `guru` / `r3tr0gradE99!` (Domain Admin;
same on both servers; console+RDP).
- `clients/prairie-schooner/tps-administrator` — built-in `Administrator` / `r3tr0gradE99#`
(Enterprise+Schema+Domain Admin; **currently DISABLED**; pw was reset to this value).
- `clients/prairie-schooner/tps-svr-dsrm` — DSRM (Directory Services Restore Mode) pw
`TpsDSRM-Recover-2026#` for TPS-SVR.
- `clients/prairie-schooner/gururmm-site-main` — pre-existing RMM enrollment key.
## Infrastructure & Servers
- Domain: `TPS.local` (NetBIOS TPS), forest/domain functional level 2016, schema now v88.
- **TPS-SERVER** (OLD, primary): 192.168.1.125 (+ .120, + 10.10.10.5 multi-homed cruft),
Server 2016 Essentials, all 5 FSMO, Enterprise Root CA `TPS-TPS-SERVER-CA`, DNS, file server,
QuickBooks host. RMM agent id `4f8dff0c-b033-4f66-9101-9350e32e0a81`.
- **TPS-SVR** (NEW, 2nd DC): 192.168.1.135/24 static, Server 2019 Standard, GC+DNS, no FSMO.
RMM agent id `799675cf-5c94-465a-b653-63e08d016526`.
- Workstations: Mel-PC (192.168.1.149, `eb2f2aeb-f56e-4fc6-8e5a-73257b95e5f0`, user erin),
TPS-JAYMI (`1fbf4e45-1b48-4c82-af85-847fc1344e12`, user jaymi), TPS-Tina
(`c8d76770-dc9e-45b7-94c7-fdcab3e105a2`, user tina). All DHCP, DNS=192.168.1.125.
- Gateway/DHCP: **SonicWall 192.168.1.1** (hands clients DNS=.125) — being replaced with UniFi.
- QuickBooks: Enterprise Server 24 (34.0.401x), company file `C:\Share\Quickbooks\schoonerQB2025.QBW`
(~3 GB), engine `QB_TPS-SERVER_34`, `.ND` `ServerIp=192.168.1.125` ServerPort 64189.
- GuruRMM API `http://172.16.3.30:3001` (vault `infrastructure/gururmm-server.sops.yaml`).
## Commands & Outputs
- Pre-seed: `robocopy C:\Share \\TPS-SVR\Share /E /COPY:DAT /DCOPY:DAT /B /R:1 /W:2 /Z /XJ /MT:16`
(detached SYSTEM task). Result: 166.55 GB / 113,971 files.
- Clean QB copy: stop `QuickBooksDB34`,`QBCFMonitorService` -> `robocopy ... schoonerQB2025.QBW
schoonerQB2025.QBW.TLG /COPY:DAT` -> restart services. SRC=DST 3044.8 MB, MATCH True.
- DC test error surfaced only via `Test-ADDSDomainControllerInstallation ... *>&1` (warning stream).
- Replication verified: `repadmin /replsummary` -> 0 fails/5 both directions.
- Static IP: `netsh interface ipv4 set address "name=Embedded LOM 1 Port 1" source=static
address=192.168.1.135 mask=255.255.255.0 gateway=192.168.1.1` (agent "restarted during
execution" = expected NIC blip; verified DHCP=Disabled, PrefixOrigin=Manual).
## Pending / Incomplete Tasks
Resume AFTER Howard swaps SonicWall -> UniFi gateway (regains DHCP/DNS control):
1. Transfer 5 FSMO roles to TPS-SVR (briefly re-enable Administrator, then re-disable).
2. Migrate Enterprise Root CA `TPS-TPS-SERVER-CA` — TEST `Backup-CARoleService` key export FIRST
(key non-exportable / Microsoft Software KSP); it still auto-enrolls DC certs.
3. Configure QuickBooks Database Server Manager on TPS-SVR (host `C:\Share\Quickbooks`, service
Automatic, DELETE stale `.ND` so it regenerates with new host IP, open QB firewall ports);
recreate shares with proper ACLs (pre-seed was data-only); re-enable Shadow Copies.
4. With UniFi: point DHCP DNS at the new DC (removes forced .125 takeover); decide final DC IP
(keep .135 + name alias, or assume .125 + name); reserve the DC IP in UniFi DHCP.
5. Cutover (brief, users out of QB): stop QB on old, final `.QBW` delta copy, verify a workstation
opens QB from the new server.
6. Demote + power off TPS-SERVER; add `TPS-SERVER` DNS+SPN alias on new box; clean multi-homing
(.120 / 10.10.10.5 DNS records); remove unused DHCP/Print roles on TPS-SVR; prune legacy GPOs
(IP SEC, NeT Communication, LT Deployment).
7. **Stand up a working backup on the new server** — the original reason this started (nothing has
been backed up since 2018).
## Reference Information
- Resume checklist: `wiki/clients/the-prairie-schooner.md` (migration section).
- Syncro cid 3664974. Contact: Tina Coalter, Tucson AZ.
- GPS audit: `wiki/projects/gps-rmm-audit.md`.