Galaxy inventory

Galaxy is my five-node Proxmox VE cluster. This page is the current inventory: the physical nodes, every guest and template, and what each one is for. I read the figures back from pvesh get /cluster/resources and the guest configuration files on 2026-09-24, and the service versions from the running services on 2026-09-24 and 2026-09-25. Two later changes are folded in: the ubuntu-dev disk expansion on 2026-09-28 and the Sure deployment on 2026-10-05.

Note: The repo keeps this state in four living files: Galaxy Inventory, VMs, LXCs and Services. The ten dated inventory sets that came before them moved to the archive on 2026-09-25 and nothing is added to them.

At a glance

Item	Value (2026-09-24)
Nodes	5, all on pve-manager/9.2.11, quorum 5 of 5 votes
Compute	30 physical cores, 38 hardware threads, 114.78 GiB of usable memory
Guests	18: 12 QEMU VMs and 6 LXCs, plus 3 templates
Running	16; kali-pen and HQ-WS001 were stopped
High availability	None. ha-manager config returns nothing
Shared storage	None. Every guest lives on its own node’s disks
Empty node	green-server, which has unresolved host memory errors from 2026-09-20
Monitoring	Prometheus scraped 57 targets in 7 jobs, all up; the Wazuh manager listed 15 active remote agents

On 2026-09-24 four nodes ran kernel 7.0.14-8-pve and blue-server ran 7.0.14-15-pve, which all five had installed at the 2026-09-06 reading.

Nodes

Node	Hardware	CPU (cores / threads)	Memory (usable, used on 2026-09-24)	Storage	What runs there
grey-server	MSI MS-7C91 board, NVIDIA GeForce GTX 1080 Ti	AMD Ryzen 7 3700X (8 / 16)	64 GB installed, 62.72 GiB usable, 51.61 GiB used	1 TB NVMe (boot and local-lvm), 2 TB SATA SSD (ssd-lvm1), 2 TB HDD (hddpool-1, ZFS)	10 guests and all 3 templates: the identity plane, the security stack, the development workstations and docker-main
purple-server	Lenovo 10RRS0LN00	Intel Core i5-8500T (6 / 6)	16 GB installed, 15.46 GiB usable, 11.68 GiB used	256 GB NVMe (boot and local-lvm), 250 GB SATA SSD (ssd-lvm2, empty)	app-01, edge-01, alpha-prod-01
blue-server	Lenovo 10MUS08B00	Intel Core i5-7500T (4 / 4)	6 GB installed, 5.68 GiB usable, 2.99 GiB used	256 GB NVMe (boot and local-lvm), 500 GB HDD (unused)	ansible-01, monitor-01, docker-network, docker-blue
red-server	Lenovo 10RRS0LN00	Intel Core i5-8500T (6 / 6)	16 GB installed, 15.46 GiB usable, 3.29 GiB used	256 GB NVMe (boot and local-lvm), 1 TB HDD (media data)	media-01
green-server	Lenovo 10RRS0LN00	Intel Core i5-8500T (6 / 6)	16 GB installed, 15.46 GiB usable, 1.91 GiB used	256 GB NVMe (boot and an empty local-lvm), 320 GB HDD (failed, not used)	Nothing. It holds no guests until a full memory test passes

grey-server, blue-server and red-server draw power from UPS-02; the records do not report a UPS for the other two. The four Lenovo nodes use their integrated Intel graphics, and Red’s UHD Graphics 630 is the render device media-01 passes through for Jellyfin Quick Sync.

A few hardware facts the table cannot hold:

• Purple’s boot drive was replaced on 2026-07-25. The Samsung NVMe it shipped with wore out at 169% of its rated write endurance, so I cloned it onto a Toshiba THNSF5256GPUK and added the 850 EVO on SATA at the same time.
• Blue’s 500 GB HDD passed a full extended SMART read on 2026-07-31 and is blank. It is available if I want to assign it.
• Green’s 320 GB HDD failed its extended test at 60 percent with a read failure and two pending sectors. It stays installed only as failed hardware, with no partition, filesystem or Proxmox storage on it.
• Green’s memory failed an online test on 2026-09-20 with 25 failure lines, and a locked 8 GiB test failed again six seconds after launch on 2026-09-27. Isolating the bad module or slot needs physical access.

Cluster storage

local and local-lvm exist on every node. The other three storages are restricted to one node each, so pvesm status reports them disabled when you ask any other node. That is how ssd-lvm2 was once recorded as disabled when it never was.

Node	Storage	Type	Total	Used % (2026-09-24)
grey-server	hddpool-1	zfspool	1,798.5 GiB	70.7%
grey-server	ssd-lvm1	lvmthin	1,830.9 GiB	11.0%
grey-server	local-lvm	lvmthin	793.8 GiB	30.3%
purple-server	local-lvm	lvmthin	140.9 GiB	28.4%
purple-server	ssd-lvm2	lvmthin	228.1 GiB	0.0%
blue-server	local-lvm	lvmthin	141.2 GiB	39.0%
red-server	local-lvm	lvmthin	141.2 GiB	31.9%
green-server	local-lvm	lvmthin	141.2 GiB	0.0%

After the 2026-09-28 expansion of ubuntu-dev, Grey’s local-lvm stood at 32.12%. Purple’s thin volumes are provisioned at 154.01 GiB against a 140.87 GiB pool, so any new allocation there needs a capacity check first.

Warning: hddpool-1 read 82.46% on 2026-09-06 and 70.7% on 2026-09-24. No record explains the roughly 200 GiB it gave back, and finding it is an open item in the Galaxy backlog.

Virtual machines

VMID	Name	Node	State	OS	vCPU	Memory	Disk	VLAN	Purpose
102	kali-pen	grey-server	stopped	Kali Linux 2026.2	6	8 GiB	100G	40	Penetration-testing VM, rebuilt on 2026-08-26 under a new VMID
103	win11-dev	grey-server	running	Windows 11 Pro 25H2	4	8 GiB	120G	40	Standalone Windows development workstation with Secure Boot and TPM 2.0; activation still open
105	ubuntu-dev	grey-server	running	Ubuntu 26.04.1 LTS, GNOME 50	6	12 GiB	230G	40	My development workstation and a Docker host; also the Samba file server for my own machines
109	splunk-siem	grey-server	running	Rocky Linux 10.2	6	12 GiB	150G	72	SIEM: Splunk Enterprise 10.4.0, Enterprise Security 8.5.1 and SC4S for UniFi CEF
116	app-01	purple-server	running	Debian 13 (trixie)	4	8 GiB max, 4 GiB min	64G	80	App platform: Coolify 4.3.23 with its Traefik 3.7.12 proxy
121	edge-01	purple-server	running	Debian 13 (trixie)	2	4 GiB max, 2 GiB min	30G	30	Edge host for the public services
200	security-01	grey-server	running	Ubuntu 24.04.4 LTS	4	10 GiB max, 8 GiB min	100G	72	Wazuh 4.14.8 manager, indexer and dashboard
301	HQ-DC01	grey-server	running	Windows Server 2025 Standard	4	4 GiB	80G	65	First domain controller: all five operations master roles, global catalog, AD-integrated DNS
302	HQ-DC02	grey-server	running	Windows Server 2025 Standard	4	4 GiB	80G	65	Second domain controller: global catalog and AD-integrated DNS
303	HQ-MGT01	grey-server	running	Windows Server 2025 Standard	2	6 GiB	100G	65	Management and hybrid identity: Windows Admin Center, Entra provisioning agent, Azure Arc agent
310	HQ-WS001	grey-server	stopped	Windows 11 Pro 25H2	4	8 GiB	80G	65	Domain-joined test workstation for tiered policy and Windows LAPS
401	alpha-prod-01	purple-server	running	Debian 13 (trixie)	6	4 GiB max, 2 GiB min	60G	80	Voice services: two TeamSpeak 3 Server 3.13.8 instances, TS3 Manager and a reachability collector

Every VM uses OVMF firmware, the virtio-scsi-single controller and the QEMU guest agent. All five Windows VMs carry TPM 2.0. None is an HA resource.

LXC containers

All six containers are unprivileged and were running on 2026-09-24. Together they hold 13 vCPUs, 27 GiB of memory and 8.5 GiB of swap.

CTID	Name	Node	OS	vCPU	Memory	Root disk	VLAN	Purpose
100	ansible-01	blue-server	Debian 13 (trixie)	1	1 GiB	16G	40	Automation and node provisioning: Ansible 14.2.0, Semaphore 2.18.27, the Galaxy PXE service and TFTP
104	monitor-01	blue-server	Debian 13 (trixie)	2	2 GiB	20G	73	Monitoring: Prometheus 3.14.0, Grafana 13.2.2, the exporters, the Discord alert bot and PeaNUT 6.0.0
107	docker-network	blue-server	Debian 13 (trixie)	2	2 GiB	32G	85	Network access: Nginx Proxy Manager 2.15.1 and NetBird 0.79.0
108	docker-blue	blue-server	Debian 13 (trixie)	2	2 GiB	15G	40	Remote access and integrations: RustDesk 1.1.16 and the MeshCentral 1.2.6 pilot
110	docker-main	grey-server	Debian 12 (bookworm)	4	16 GiB	100G, plus 2,900G /data on hddpool-1	40	Docker apps, with the GTX 1080 Ti passed through: Immich, Forgejo, BookOrbit, Homarr, Sure, Dockhand, Ollama, Open WebUI, App Portal, CLI Proxy API
842	media-01	red-server	Debian 13 (trixie)	2	4 GiB	100G, plus 916G usable on the HDD bind mount	40	Media: Jellyfin 12.1.0, Seerr, Sonarr, Radarr, Prowlarr, FlareSolverr and qBittorrent behind Gluetun and Proton VPN

A few details worth knowing:

• docker-network and docker-blue are no longer HA resources. After the 2026-07-20 stranding incident I pinned both to blue-server with a strict node-affinity rule. Both resources and the rule were present on 2026-09-06 and gone by 2026-09-23, with no record of the removal. A Blue outage now stops both until Blue returns.
• media-01 refuses to start without its HDD. The data directory exists only on the ext4 filesystem the host automounts, so the stack fails closed instead of writing media onto its root volume. It also gets Red’s render device for Jellyfin Quick Sync and /dev/net/tun for the Gluetun WireGuard tunnel.
• docker-main sits outside the standard account model. It is root-login only, keyed, and is the one guest still on Debian 12.
• Nineteen containers ran on docker-main on 2026-10-05 after Sure added four. The internal documentation site there has been stopped since 2026-09-22.

Templates

VMID	Name	OS	vCPU	Memory	Disk	VLAN	Notes
101	debian13-template	Debian 13	4	4 GiB	60G	40	Base for new Debian guests
300	ws2025-template	Windows Server 2025 Standard	4	4 GiB	80G	65	Built 2026-09-08 as the source for the three Windows Server guests; QEMU agent, OpenSSH and virtio drivers, sysprepped
9000	ubuntu-cloud-template	Ubuntu 24.04.4 LTS	2	2 GiB	20G	80	Cloud-init template

All three live on grey-server in ssd-lvm1.

Warning: Sysprep does not regenerate SSH host keys, so every clone of ws2025-template presents the template’s host key fingerprint until the keys under C:\ProgramData\ssh are deleted and sshd restarts. Any guest cloned from the Debian or Ubuntu template also arrives with root locked, which matters because the sudo policy on the fleet points the sudo prompt at root’s password. The account play has to run on a new host before the sudo play does.

Workload placement in one view

grey-server    HQ-DC01  HQ-DC02  HQ-MGT01  HQ-WS001 (stopped)
               splunk-siem  security-01
               win11-dev  ubuntu-dev  kali-pen (stopped)
               docker-main (CT 110)
               templates 101, 300, 9000
purple-server  app-01  edge-01  alpha-prod-01
blue-server    ansible-01  monitor-01  docker-network  docker-blue
red-server     media-01
green-server   (empty)

Most of that placement is recent. Between 2026-09-11 and 2026-09-12 I moved app-01, edge-01 and alpha-prod-01 from Grey to Purple’s NVMe and ansible-01 from Grey to Blue. On 2026-09-23 win11-dev moved from Green to Grey after Green’s memory test failures. The change history has each move.

Monitoring and agent coverage

• Prometheus covers every running Linux guest with node_exporter and the Docker hosts with cAdvisor. All five nodes export node metrics, and UPS-02 on grey-server reports through NUT. UPS-01 has had no data path since its cable was disconnected on 2026-08-28.
• Wazuh listed 15 active remote agents on 2026-09-24: the five Proxmox nodes and ten Linux guests. splunk-siem and the Windows guests run no agent.
• Dockhand manages docker-main through its local socket and six more Docker hosts through Hawser agents. app-01 has no Hawser agent, so Coolify’s containers stay outside Dockhand.

Retired guests

Retired guests stay in the inventory files as a list, and their full records move to the archive. The retired guests page summarizes them.

ID	Name	Removed
CT 104	ai-alpha-01 (OpenClaw)	Already gone when checked on 2026-07-25
VM 111	fedora-dev	Confirmed absent on 2026-08-08
CT 105	ai-bravo-02 (TNIO AI Bot)	Deleted with its root volume on 2026-08-09
VM 102	debian-dev	Destroyed 2026-08-14 after ubuntu-dev replaced it
VM 122	kasm-01	Destroyed with all volumes on 2026-08-19
VM 117	supabase-01	Confirmed deleted on 2026-08-20
VM 106	kali-pen (first build)	Destroyed 2026-08-26 and rebuilt as VM 102
CT 123	game-01	Deleted with its game data on 2026-09-12

VMIDs get reused. VM 102 was debian-dev until 2026-08-14 and has been kali-pen since 2026-08-26, and CT 104 was ai-alpha-01 before it was monitor-01.