# Fleet Proxmox host inventory — 2026-10-02 (snapshot)

Requested by Prime via Miranda for PVE 8→9 planning. A point-in-time read; re-read live before acting. ESH cluster detail: `docs/runbooks/esh-pve-cluster-pve9-upgrade-plan.md`.

FLEET PROXMOX HOST INVENTORY — all sites. Captured 2026-10-02 ~19:45 PT, live (ssh + pvesh + guest agent/pct exec), read-only.

SITE SUMMARY
| Site | Proxmox hosts | PVE major (exact) | Notes |
|---|---|---|---|
| ANA (Anaheim colo) | pfi-pve — read live | **8** (8.3.5) | Dell PowerEdge R750xs, standalone. sfsrv-ana (SureFire client's PVE host) NOT read — see below |
| NH3 | nh3-pve (node name nh3-vmhost) — read live | **8** (8.4.1) | MS-01, standalone. nh3-pve-2 (new MS-03) is mid-install with the **9**.2 ISO, not on the network yet |
| FV (Fountain Valley) | **none** | n/a | fv-ml1 is bare-metal Debian 13 (GPU host), not Proxmox |
| IRV (Irvine) | **none** | n/a | irv-ml1 is bare-metal Debian 12 (GPU host), not Proxmox |
| ESH (home lab) | pve + esh-nas-pve, 2-node cluster `esh-pve-cluster` | **8** (8.4.20) | See the earlier ESH block (10 guests). ⚠ its 175 T `tank` pool is DEGRADED |

HOST: pfi-pve (ANA) — 10.250.250.31
- Hardware: Dell PowerEdge R750xs, Intel Xeon Silver 4310 (48 threads), 188 GiB RAM. Uptime 3 weeks.
- PVE 8.3.5 (needs 8.4.x before any 9 upgrade), kernel 6.8.12-8, Debian 12. Standalone (no cluster).
- Root: ext4 on LVM (31G free). Storage: ospool (ZFS 10.9T, 20% used — most guest disks), NASPool (ZFS 32.7T, as dir 'naspool-vmstorage', 1.5% used), local-lvm (lvmthin 130G), pbs-ana. Both zpools ONLINE.
- Backups: vzdump 03:00 daily, ALL except 100/109/114, snapshot mode -> pbs-ana (keep last 5 + 1 daily/weekly/monthly/yearly); separate 22:00 job for CT 109 only.
- Guests: 14 guests (14 running), 117 vCPU / 106.508 GiB RAM allocated
| VMID | Type | Name | Status | vCPU | RAM | Disk | OS (live) | Role | PBS |
|---|---|---|---|---|---|---|---|---|---|
| 100 | VM | pbs-ana | running | 4 | 8G | scsi0 32G local-lvm(lvmthin) | Debian GNU/Linux 12 (bookworm) | PBS (pbs-ana) — backup target for ALL sites | NONE |
| 101 | VM | PFI-ANA-DC | running | 12 | 24G | scsi0 240G ospool(zfspool) | Windows Server 2022 Datacenter | Windows AD domain controller | 8 backups, newest 2026-10-02 |
| 102 | VM | PFI-ANA-Docker | running | 8 | 16G | scsi0 250G ospool(zfspool) | Debian GNU/Linux 12 (bookworm) | ana-docker: LiteLLM gateway, traefik, gitea, vaultwarden, Beszel/Dozzle hubs, RustDesk | 8 backups, newest 2026-10-02 |
| 103 | VM | PFI-SlaveBot | running | 8 | 8G | ide0 256G ospool(zfspool) | Windows 10 IoT Enterprise LTSC 2021 | Windows 10 IoT utility VM | 8 backups, newest 2026-10-02 |
| 104 | VM | PFI-Mongo | running | 8 | 8G | scsi0 256G ospool(zfspool) | Debian GNU/Linux 12 (bookworm) | MongoDB | 8 backups, newest 2026-10-02 |
| 105 | VM | PFI-Postgres | running | 16 | 8G | scsi0 80G ospool(zfspool) | Debian GNU/Linux 12 (bookworm) | shared Postgres (vaultwarden/gitea/paperless) | 8 backups, newest 2026-10-02 |
| 106 | VM | corviduo-dev | running | 8 | 8G | scsi0 80G ospool(zfspool) | Debian GNU/Linux 13 (trixie) | Worldtree dev VM (demo/personal/pinned deployments) | 8 backups, newest 2026-10-02 |
| 107 | VM | PFI-Pteradactyl | running | 8 | 8G | scsi0 256G ospool(zfspool) | Debian GNU/Linux 12 (bookworm) | Pterodactyl game panel | 8 backups, newest 2026-10-02 |
| 109 | CT | ana-nas | running | 4 | 2G | rootfs 80G ospool(zfspool) | Debian GNU/Linux 12 (bookworm) | ana-nas: NFS/SMB storage server (storage SPOF) | 28 backups, newest 2026-10-02 |
| 110 | VM | PFI-ANA-Webhost | running | 16 | 4G | scsi0 250G ospool(zfspool) | Debian GNU/Linux 11 (bullseye) | web host (Debian 11) | 8 backups, newest 2026-10-02 |
| 111 | VM | pfi-tacticalrmm | running | 16 | 8G | scsi0 256G ospool(zfspool) | Debian GNU/Linux 12 (bookworm) | TacticalRMM + MeshCentral | 8 backups, newest 2026-10-02 |
| 112 | CT | ana-filebot | running | 4 | 2G | rootfs 80G ospool(zfspool) | Debian GNU/Linux 12 (bookworm) | file-task automation | 8 backups, newest 2026-10-02 |
| 113 | CT | ana-wg | running | 4 | 2G | rootfs 8G ospool(zfspool) | Debian GNU/Linux 12 (bookworm) | WireGuard | 8 backups, newest 2026-10-02 |
| 114 | CT | ana-scale | running | 1 | 0.5G | rootfs 8G ospool(zfspool) | Debian GNU/Linux 12 (bookworm) | ANA headscale subnet router (mesh) | 1 backup, newest 2026-09-06 |

HOST: nh3-pve (NH3) — 10.100.250.60 (Proxmox node name nh3-vmhost)
- Hardware: Minisforum MS-01, i9-13900H (20 threads), 62 GiB RAM, RTX 2000E Ada (for CT 109). Uptime 1 week. Intel AMT phones home to MeshCentral.
- PVE 8.4.1, kernel 6.8.12-43, Debian 12. Standalone (no cluster).
- Root: ZFS rpool (952G, 667G free). Storage: local-zfs (912G, 27% used — all guest disks), pfi-nh3-nas (NFS, 42.9T, 77% used), pbs-ana.
- Backups: vzdump 21:00 daily, ALL except 107/109, snapshot mode -> pbs-ana (same retention); pbs-nh3 (VM 105) syncs pbs-ana as the DR mirror.
- Guests: 10 guests (8 running), 58 vCPU / 83 GiB RAM allocated
| VMID | Type | Name | Status | vCPU | RAM | Disk | OS (live) | Role | PBS |
|---|---|---|---|---|---|---|---|---|---|
| 100 | VM | nh3-docker1 | running | 8 | 8G | scsi0 128G local-zfs(zfspool) | Debian GNU/Linux 12 (bookworm) | nh3-docker: althing post office, SearXNG, AdGuard NH3, albok-service | 8 backups, newest 2026-10-02 |
| 101 | VM | nh3-extdev | running | 8 | 8G | scsi0 256G local-zfs(zfspool) | Debian GNU/Linux 13 (trixie) | manager / external-dev box | 8 backups, newest 2026-10-02 |
| 102 | VM | nh3-dev | running | 16 | 28G | scsi0 378G local-zfs(zfspool) | Debian GNU/Linux 12 (bookworm) | nh3-dev: live Claude sessions, svos/hermes (Miranda channel), Booth, fleet TLS caddy | 8 backups, newest 2026-10-02 |
| 103 | CT | nh3-wg | running | 4 | 2G | rootfs 40G local-zfs(zfspool) | Ubuntu 22.04.5 LTS | nh3-wg (WireGuard) | 8 backups, newest 2026-10-02 |
| 104 | VM | nh3-laser | stopped | 8 | 8G | ide0 120G local-zfs(zfspool) | win11 | Windows 11 (laser) — stopped | 8 backups, newest 2026-10-02 |
| 105 | VM | pbs-nh3 | running | 4 | 8G | scsi0 32G local-zfs(zfspool) | Debian GNU/Linux 12 (bookworm) | PBS DR mirror (pbs-nh3) | 8 backups, newest 2026-10-02 |
| 106 | CT | nh3-headscale | running | 1 | 0.5G | rootfs 8G local-zfs(zfspool) | Debian GNU/Linux 12 (bookworm) | headscale control server (mesh) | 7 backups, newest 2026-10-02 |
| 107 | CT | nh3-scale | running | 1 | 0.5G | rootfs 8G local-zfs(zfspool) | Debian GNU/Linux 12 (bookworm) | NH3 headscale subnet router (mesh) | NONE |
| 108 | VM | opnsense-lab | stopped | 2 | 4G | scsi0 32G local-zfs(zfspool) | other | OPNsense lab — stopped | 7 backups, newest 2026-10-02 |
| 109 | CT | nh3-ml1 | running | 6 | 16G | rootfs 80G local-zfs(zfspool) | Debian GNU/Linux 12 (bookworm) | GPU CT: TEI embed/rerank (twin of esh-ml1) | NONE |

HOST: nh3-pve-2 (NH3) — new Minisforum MS-03 (Core Ultra 9 Panther Lake, 64 GB). PVE 9.2 installed from the ISO but onto the wrong NVMe; reinstall planned on site 2026-10-03. Host not on the network yet; its AMT (21.0.6) is at 10.100.250.63 and phones home to MeshCentral. No guests.

HOST: sfsrv-ana (ANA) — 10.250.250.115, SureFire client's Proxmox host (client-owned Dell R630, PFI-managed). NOT READ: our standing rule is to coordinate before touching the SureFire hosts, so version, guests and backup posture are unknown here. Fleet docs record backup coverage there as 'status unknown'. A read-only `pveversion` + guest list needs Prime's OK.

BACKUP POSTURE ACROSS SITES (vzdump -> pbs-ana at Anaheim; pbs-nh3 mirrors it)
- Every guest is backed up nightly EXCEPT: pfi-pve 100 (pbs-ana itself — expected), 114 ana-scale (excluded; one stale backup from 2026-09-06); nh3-pve 107 nh3-scale and 109 nh3-ml1 (excluded); ESH 108 esh-scale and 110 esh-ml1 (no job covers them).
- Pattern: all three headscale subnet routers (ana-scale 114, nh3-scale 107, esh-scale 108) are excluded. Likely deliberate after a vzdump lock on esh-scale once blackholed ESH, but none of them has a current restore point. Both GPU/TEI containers (nh3-ml1 109, esh-ml1 110) are also uncovered.
- pfi-pve CT 109 (ana-nas, the Anaheim storage SPOF) has its own 22:00 job (28 restore points). Note that an older standing rule (2026-04) said not to vzdump CT 109; the dedicated job postdates it.

PVE 9 READINESS, PER HOST
- pfi-pve: 8.3.5 -> latest 8.4 first. Hosts pbs-ana (every site's backup target), the LiteLLM gateway (ana-docker), TacticalRMM, shared Postgres and the ana-nas storage SPOF: a reboot is a fleet-wide event. Standalone, so no quorum concerns.
- nh3-pve: 8.4.1 -> latest 8.4 first. Hosts nh3-dev (all agent sessions, Miranda's svos/hermes channel), nh3-docker (althing post office, albok, AdGuard NH3), headscale + the NH3 mesh router, pbs-nh3: a reboot takes down the fleet bus and Miranda's channel. Its AMT now phones home (remote recovery exists).
- nh3-pve-2: already 9 (fresh install).
- ESH cluster: 8.4.20; plan delivered separately (pool repair first).