Skip to content

Full infrastructure inventory — October 15, 2026

Status: dated, read-only inventory. This replaces the September hardware snapshot in Explore the lab.

Collection: approximately 04:00–04:10 UTC; LXC configuration checked at 04:12 UTC. Kubernetes, Talos and host queries are separate samples, not an atomic snapshot.

Coverage and limits

  • Seven physical machines: five Proxmox hosts, TrueNAS and the management Pi.
  • Seven Proxmox VMs: six running Talos VMs and one stopped Kali VM. Elite also runs LXC 101, Proxmox Datacenter Manager.
  • One additional stopped TrueNAS VM; three running TrueNAS apps and nine running Pi containers.
  • Twenty-six visible physical drives. The missing Dell T-FORCE is excluded. NAS zvols and guest disks are excluded.
  • All six Kubernetes nodes reported Ready: Talos v1.14.2, Kubernetes v1.37.1, containerd 2.3.6.
  • All 85 PVCs were Bound. Longhorn reported 65 healthy attached volumes and seven detached volumes with unknown robustness.
  • Argo CD reported 97 Healthy applications and one Degraded application: my-apps-wyze-bridge.

All 26 visible physical drives reported passing SMART status. NAS HDD temperatures ranged from 32 to 50 °C.

A Ready node, mounted filesystem or passing SMART status does not prove application health or successful storage I/O. No benchmarks, SMART self-tests, filesystem repairs, backup restores or infrastructure changes were performed.

Changes and findings

  1. Dell storage mismatch: Proxmox lists dell-ssd-vmstore inactive; its VG and physical T-FORCE disk are absent. VM 100 still declares its 400 GiB scsi1; Talos still reports the XFS mount and Longhorn still reports free capacity. Host logs record an ATA link loss and removal of sda. This audit does not establish the cause or data integrity. PR #2946 records the T-FORCE replacement, preserving the same VM and Longhorn names.
  2. SFF control plane: the Intel S3610 now backs VM 101. One PNY SSD is an orphan PV outside any VG.
  3. Dell Apple SSD: the 220 GiB guest disk backs the separate dell-apple Longhorn path.
  4. GPU host: two RTX 3090 PCI devices are mapped into the Talos VM. Its HPE RAID1 reports both members present ([UU]).
  5. NAS: all four imported pools report ONLINE with no known data errors. The Samsung 850 EVO is visible outside imported pools. PR #2946 records its retirement after failed load tests; its current SMART pass does not reverse that finding.
  6. Elite endurance: Intel 660p reports 76% used and zero NVMe media errors. This is an endurance indicator, not a failure probability.
  7. NAS replication: the local replication task is enabled without a schedule; its last recorded completion remains May 2025. This audit does not establish any off-box backup copy.
  8. Wyze Bridge: Argo health cites ExternalSecret wyze-web-session-seed. This audit does not diagnose its cause.

Hosts

RAM below is kernel-visible capacity. Configured VM RAM appears separately.

Host Address CPU Visible RAM GiB Available RAM GiB Software
HP SFF 192.168.10.21 Intel(R) Core(TM) i5-8500 CPU @ 3.00GHz · 6 threads 62.6 10.0 pve-manager/9.2.21/4f6e0ac86f9e8c7f (running kernel: 7.0.14-20-pve)
HP Elite Mini 192.168.10.22 13th Gen Intel(R) Core(TM) i5-13500T · 20 threads 30.6 2.0 pve-manager/9.2.21/4f6e0ac86f9e8c7f (running kernel: 7.0.14-20-pve)
Threadripper 192.168.10.14 AMD Ryzen Threadripper 2950X 16-Core Processor · 32 threads 125.7 19.8 pve-manager/9.2.21/4f6e0ac86f9e8c7f (running kernel: 7.0.14-20-pve)
Dell OptiPlex 192.168.10.16 Intel(R) Core(TM) i5-8500 CPU @ 3.00GHz · 6 threads 39.0 6.7 pve-manager/9.2.21/4f6e0ac86f9e8c7f (running kernel: 7.0.14-23-pve)
Shed HP Mini 192.168.10.20 Intel(R) Core(TM) i5-8500T CPU @ 2.10GHz · 6 threads 31.1 24.0 pve-manager/9.2.21/4f6e0ac86f9e8c7f (running kernel: 7.0.14-20-pve)
Omni / DNS Pi 192.168.10.15 Cortex-A76 · 4 threads 7.7 5.7 Ubuntu 26.04 LTS
TrueNAS 192.168.10.133 Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GHz · 28 threads 377.6 283.5 TrueNAS-28.0.0-MASTER+20261012-230219

Virtual machines

VM IDs are local to each host. NAS VM 5 is separate from the six active Talos nodes.

Host VM Name State vCPU RAM MiB Node IP
sff 100 talos-prod-cluster-v2-hp-sff-workers-h76ncc running 6 40960 192.168.10.173
sff 101 talos-prod-cluster-v2-control-planes-7mxjxv running 4 12288 192.168.10.94
elite 100 talos-prod-cluster-v2-hp-elite-workers-l5zlgr running 16 26624 192.168.10.129
gpu 100 talos-prod-cluster-v2-gpu-workers-jxfkdr running 30 102400 192.168.10.187
gpu 102 kali-linux stopped 4 12000 Not collected
dell 100 talos-prod-cluster-v2-dell-workers-2jfrb8 running 6 30720 192.168.10.84
shed 100 talos-prod-cluster-v2-hp-micro-workers-pstbbd running 4 12288 192.168.10.128
nas 5 talos_prod_control_03 STOPPED; autostart false 2 8000 Not collected

LXC guest

Elite runs proxmox-datacenter-manager as unprivileged Debian container 101 with autostart enabled. It has 2 vCPU, 2,048 MiB RAM, 512 MiB swap and a 10 GiB root filesystem. The backing volume is hp-elite-vmstore:vm-101-disk-0, on the Intel 660p. Its network uses DHCP on vmbr0; the guest address was not collected.

The stopped NAS VM uses /dev/zvol/Backup10T/virtual-machine/talos_prod_control_03-f06my4. Its installation ISO is also on Backup10T; sanitized paths appear in the inventory JSON.

Physical drives

Capacity uses decimal GB. Device names can change after reconnecting hardware.

Host Device Model GB Path / role SMART and health
sff sda PNY CS900 1TB SSD 1000.2 No volume group SMART overall passed; 33 °C at collection.
sff sdb INTEL SSDSC2BX400G4R 400.1 hp-sff-cp → VM 101 scsi0 (100 GiB) SMART overall passed; 30 °C at collection.
sff sdc PNY CS900 1TB SSD 1000.2 pve / hp-prodesk-vmstore → VM 100 scsi0 + scsi1 SMART overall passed; 33 °C at collection.
elite nvme0n1 INTEL SSDPEKNW512G8 512.1 hp-elite-vmstore → VM 100 scsi1 (440 GiB); LXC 101 rootfs (10 GiB) SMART overall passed; 35 °C at collection. NVMe endurance used: 76%; media errors: 0.
elite nvme1n1 WDC PC SN530 SDBPMPZ-256G-1101 256.1 pve / local-lvm → VM 100 scsi0 (128 GiB) SMART overall passed; 44 °C at collection. NVMe endurance used: 3%; media errors: 0.
gpu sda PNY CS900 1TB SSD 1000.2 pve / tr-sda-vmstore → VM 100 scsi4 (800 GiB) SMART overall passed; 33 °C at collection.
gpu sdb MK000480GWCEV 480.1 md0 → ssd-ent → VM 100 scsi3 (440 GiB) SMART overall passed; 18 °C at collection.
gpu sdc MK000480GWCEV 480.1 md0 → ssd-ent → VM 100 scsi3 (440 GiB) SMART overall passed; 24 °C at collection.
gpu nvme0n1 EDILOCA EN605 512GB 512.1 nvme0-vmstore → VM 100 scsi0 (16 GiB) + scsi1 (434 GiB) SMART overall passed; 54 °C at collection. NVMe endurance used: 16%; media errors: 0.
gpu nvme1n1 EDILOCA EN605 512GB 512.1 nvme1-vmstore → VM 100 scsi2 (450 GiB) SMART overall passed; 54 °C at collection. NVMe endurance used: 20%; media errors: 0.
dell sdb R9M2SH6Q240GQF72C 240.1 pve / local-lvm → VM 100 scsi0 (128 GiB) SMART overall passed; 31 °C at collection.
dell sdc APPLE SSD SM0256G 251.0 dell-apple-vmstore → VM 100 scsi2 (220 GiB) SMART overall passed; 29 °C at collection.
shed sda PNY CS900 1TB SSD 1000.2 hp-ssd-vmstore → VM 100 scsi1 (850 GiB) SMART overall passed; 33 °C at collection.
shed nvme0n1 SK hynix BC501 HFM256GDJTNG-8310A 256.1 pve / local-lvm → VM 100 scsi0 (128 GiB) SMART overall passed; 41 °C at collection. NVMe endurance used: 6%; media errors: 0.
pi nvme0n1 Patriot M.2 P300 256GB 256.1 Ubuntu root; Docker services SMART overall passed; 32 °C at collection. NVMe endurance used: 13%; media errors: 0.
nas sda HP SSD S700 500GB 500.1 ai-pool SMART overall passed; 44 °C at collection. Lifetime CRC counter: 816; this alone does not prove current errors.
nas sdb Samsung SSD 860 EVO 1TB 1000.2 ai-pool SMART overall passed; 32 °C at collection. Lifetime CRC counter: 4; this alone does not prove current errors.
nas sdc HUH721010AL4200 10000.8 BigTank mirror-0 SMART overall passed; 37 °C at collection.
nas sdd HUH721010AL4200 10000.8 BigTank mirror-1 SMART overall passed; 50 °C at collection.
nas sde T-FORCE 512GB 512.1 boot-pool mirror SMART overall passed; 33 °C at collection.
nas sdf HUH721010AL4200 10000.8 BigTank mirror-0 SMART overall passed; 45 °C at collection.
nas sdg P3-512 512.1 ai-pool SMART overall passed; 30 °C at collection.
nas sdh HUH721010AL4200 10000.8 BigTank mirror-1 SMART overall passed; 50 °C at collection.
nas sdi MK000480GWCEV 480.1 boot-pool mirror SMART overall passed; 19 °C at collection.
nas sdj ST10000NM0096 10000.8 Backup10T single disk SMART overall passed; 32 °C at collection.
nas sdk Samsung SSD 850 EVO 500GB 500.1 No imported pool membership observed SMART overall passed; 31 °C at collection. Lifetime CRC counter: 8162; this alone does not prove current errors. Samsung 850 EVO retired from Dell in PR #2946; SMART pass does not reverse the recorded load-test failure.

VM disk paths

These are declared Proxmox attachments. The Dell inactive backing store is an explicit exception to availability.

Host / VM Slot Backing Size
sff / 100 scsi0 hp-prodesk-vmstore:vm-100-disk-0 128 GiB
sff / 100 scsi1 hp-prodesk-vmstore:vm-100-disk-1 690 GiB
sff / 101 scsi0 hp-sff-cp-vmstore:vm-101-disk-0 100 GiB
elite / 100 scsi0 local-lvm:vm-100-disk-0 128 GiB
elite / 100 scsi1 hp-elite-vmstore:vm-100-disk-0 440 GiB
gpu / 100 scsi0 nvme0-vmstore:vm-100-disk-1 16 GiB
gpu / 100 scsi1 nvme0-vmstore:vm-100-disk-2 434 GiB
gpu / 100 scsi2 nvme1-vmstore:vm-100-disk-0 450 GiB
gpu / 100 scsi3 ssd-ent:vm-100-disk-0 440 GiB
gpu / 100 scsi4 tr-sda-vmstore:vm-100-disk-0 800 GiB
gpu / 102 scsi0 truenas-smb:102/vm-102-disk-0.raw 305 GiB
dell / 100 scsi0 local-lvm:vm-100-disk-0 128 GiB
dell / 100 scsi1 dell-ssd-vmstore:vm-100-disk-0 400 GiB
dell / 100 scsi2 dell-apple-vmstore:vm-100-disk-0 220 GiB
shed / 100 scsi0 local-lvm:vm-100-disk-0 128 GiB
shed / 100 scsi1 hp-ssd-vmstore:vm-100-disk-0 850 GiB

Proxmox storage

For thick LVM, allocated capacity is not guest filesystem usage. Thin-pool usage and NAS share capacity use different accounting.

Host Store Type Status Total GiB Used / allocated GiB Available GiB
sff hp-prodesk-vmstore lvm active 930.0 922.0 8.0
sff hp-sff-cp-vmstore lvm active 372.6 100.0 272.6
sff local dir active 93.9 13.1 76.0
elite hp-elite-vmstore lvm active 476.9 450.0 26.9
elite local dir active 67.7 12.7 51.5
elite local-lvm lvmthin active 141.2 59.7 81.6
gpu local dir active 93.9 40.9 48.4
gpu nvme0-vmstore lvmthin active 467.3 230.4 236.9
gpu nvme1-vmstore lvmthin active 467.3 29.0 438.3
gpu ssd-ent lvm active 447.0 440.0 7.0
gpu tr-sda-vmstore lvm active 930.5 904.0 26.5
gpu truenas-smb cifs active 8468.1 508.2 7959.9
dell dell-apple-vmstore lvm active 233.8 220.0 13.8
dell dell-ssd-vmstore lvm inactive 0.0 0.0 0.0
dell local dir active 66.6 15.4 47.8
dell local-lvm lvmthin active 140.0 92.2 47.8
shed hp-ssd-vmstore lvm active 931.5 850.0 81.5
shed local dir active 67.7 13.7 50.6
shed local-lvm lvmthin active 153.2 14.4 138.8

Longhorn disks

Scheduled bytes are requested replica capacity, not bytes used. They can exceed filesystem capacity under configured overprovisioning. Dell values below are controller observations despite the missing physical backing store.

Node Disk Scheduling Capacity GiB Free GiB Scheduled GiB
talos-prod-cluster-v2-dell-workers-2jfrb8 dell-apple True 219.9 212.0 0.0
talos-prod-cluster-v2-dell-workers-2jfrb8 dell-ssd True 399.8 392.1 0.0
talos-prod-cluster-v2-dell-workers-2jfrb8 talos-ephemeral False 125.8 69.4 0.0
talos-prod-cluster-v2-gpu-workers-jxfkdr gpu-bulk True 799.6 750.3 530.1
talos-prod-cluster-v2-gpu-workers-jxfkdr ssd-flash True 439.9 298.2 482.0
talos-prod-cluster-v2-gpu-workers-jxfkdr talos-ephemeral False 433.8 235.9 0.0
talos-prod-cluster-v2-hp-elite-workers-l5zlgr hp-elite-nvme True 439.8 421.4 121.0
talos-prod-cluster-v2-hp-elite-workers-l5zlgr talos-ephemeral False 125.8 74.3 0.0
talos-prod-cluster-v2-hp-micro-workers-pstbbd hp-micro-ssd False 849.6 833.2 0.0
talos-prod-cluster-v2-hp-micro-workers-pstbbd talos-ephemeral False 125.8 109.9 0.0
talos-prod-cluster-v2-hp-sff-workers-h76ncc hp-sff-ssd True 689.7 572.2 376.0
talos-prod-cluster-v2-hp-sff-workers-h76ncc talos-ephemeral False 125.8 82.3 0.0

TrueNAS and management services

Pool Size Allocated Layout
Backup10T 9.08 TiB 68% One 10 TB disk
BigTank 18.16 TiB 56% Two mirrors · four 10 TB disks
ai-pool 1.82 TiB 76% Three SSDs striped
boot-pool 0.22 TiB 25% T-FORCE + HPE mirror

ai-pool and Backup10T have no member-disk redundancy. Pool state does not prove backup freshness. The NAS exposed eight enabled NFS shares. The dataset inventory contains 203 datasets and zvols. ARC held 83.3 GiB at collection; cache size varies with workload and uptime. Share definitions and dataset usage were inspected without reading application files.

Host Service Version / image Observed state
pi proxcenter-frontend ghcr.io/adminsyspro/proxcenter-frontend:1.4.10-arm64-20261001 Up 13 days (healthy)
pi proxcenter-postgres postgres:16.15-alpine3.24 Up 13 days (healthy)
pi omni ghcr.io/siderolabs/omni:v1.12.3 Up 13 days
pi technitium-dns technitium/dns-server:15.5.1 Up 13 days
pi omni-infra-provider-proxmox-hp-micro ghcr.io/siderolabs/omni-infra-provider-proxmox:v0.3.0 Up 13 days
pi omni-infra-provider-proxmox-hp-elite ghcr.io/siderolabs/omni-infra-provider-proxmox:v0.3.0 Up 13 days
pi omni-infra-provider-proxmox-hp-sff ghcr.io/siderolabs/omni-infra-provider-proxmox:v0.3.0 Up 13 days
pi omni-infra-provider-proxmox-dell ghcr.io/siderolabs/omni-infra-provider-proxmox:v0.3.0 Up 13 days
pi omni-infra-provider-proxmox-threadripper ghcr.io/siderolabs/omni-infra-provider-proxmox:v0.3.0 Up 13 days
nas tailscale 1.4.16 RUNNING
nas rustfs 1.1.31 RUNNING
nas physical-monitoring 1.0.0 RUNNING

Evidence and refresh method

Host evidence: lsblk, lscpu, /proc/meminfo, pveversion, qm list, filtered qm config, pct list, pvesm status, LVM reports and smartctl -a. NAS evidence: zpool status, zpool list, zfs list, ARC counters, and read-only middleware queries for VMs, apps, shares and replication. Cluster evidence: Kubernetes node, pod, PVC, PV, Argo CD and Longhorn APIs; Talos disks, discoveredvolumes and mountstatus resources.

Use the README service-account access procedure before Talos queries. Use the Proxmox storage reference for storage accounting and commands. Raw responses remain private under /tmp/talos-inventory-20261015; published files omit credentials, disk serials and raw middleware configurations.

Historical benchmarks remain dated September 20 and October 11. They were not rerun or relabelled as current measurements.