sites: Inventar und Zielbild an echte Knoten (pve44–pve61) angepasst; ADR-0001/0002/0006 ergänzt
ci / test (push) Successful in 25s
ci / test (push) Successful in 25s
Aufbauplan Task 0.4: - inventory.yaml: pve44/46/48/52/61 und GODE-PC-05/06 (Ada, Windows, cpu_roles: false) - proxmox.yaml: eine GPU-VM pro Knoten mit allen Karten (D2), local-lvm/tank (D1 B), 192.168.100.0/24 ohne VLAN, Gast-IPs .220–.239 (ANNAHME bis Task 0.2) - registry.yaml/prometheus.yml: ops-Labels auf gpu-pve44/46/48, Ada ohne pve_node/vmid - tofu: dynamische hostpci, Speicher je Gast; Ansible: pve_gpu_hosts/pve_kiosk, Replikation auf mehrere Ziele - Placement: cpu_roles=false schließt Arbeitsplätze/Kiosk von CPU-Rollen aus - PLAN §1.3/§1.4/§9 und ADR-Nachträge (D1 entschieden, D2/D3 Bestätigung offen) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Claude Opus 5.5
+34
-21
@@ -1,31 +1,44 @@
|
||||
# Proxmox-Zielbild für "knaster" (Eingabe für platform/proxmox/tofu und ansible).
|
||||
# ANNAHMEN (Hosts, VM-IDs, IPs, PCI-Adressen) bis zum Inventarlauf – siehe docs/IMPLEMENTATION.md.
|
||||
# Grundlage: Aufbauplan docs/superpowers/plans/2026-10-12-aufbau-cluster-knaster.md (D1 B, D2, D5).
|
||||
# ANNAHME bis Task 0.2: Gast-IPs aus dem vorgeschlagenen Block .220–.239 (DHCP-Ausschluss noch nicht geprüft).
|
||||
# ANNAHME bis Task 3.1: PCI-Pfade/IOMMU-Gruppen wie auf pve48 geprüft; auf jedem Knoten mit `lspci -nn` bestätigen.
|
||||
cluster:
|
||||
endpoint: https://pve1:8006/
|
||||
storage_vm: tank # ZFS-Pool auf SSD 2+3 (ADR-0002)
|
||||
endpoint: https://192.168.100.210:8006/ # pve44, IP-Literal (Global Constraints)
|
||||
storage_default: local-lvm # D1 B: GPU-VMs auf der System-SSD; HA-Gäste auf `tank` (pve44/48/52)
|
||||
storage_iso: local
|
||||
template_vmid: 9000 # Ubuntu 24.04 Cloud-Image-Template (pro Node vorhanden)
|
||||
template_vmid: 9000 # Ubuntu 24.04 Cloud-Image-Template (pro GPU-Knoten vorhanden)
|
||||
lxc_template: "local:vztmpl/debian-13-standard_13.1-2_amd64.tar.zst" # ANNAHME: Dateiname nach `pveam update` prüfen
|
||||
bridge: vmbr0
|
||||
vlan_svc: 20
|
||||
gateway: 10.20.0.1
|
||||
dns: 10.20.0.1
|
||||
vlan_svc: null # flaches Netz, vorerst kein VLAN (D5)
|
||||
gateway: 192.168.100.1
|
||||
dns: 192.168.100.120
|
||||
ssh_public_key_file: ~/.ssh/id_ed25519.pub
|
||||
|
||||
pci_mappings: # Proxmox Resource Mappings (Datacenter → Resource Mappings → PCI)
|
||||
gpu-ada-1: {node: pve-ada, path: "0000:01:00", id: "10de:27b2", subsystem_id: "", iommu_group: 0}
|
||||
gpu-ada-2: {node: pve-ada, path: "0000:02:00", id: "10de:27b2", subsystem_id: "", iommu_group: 0}
|
||||
gpu-p2000-1: {node: pve3, path: "0000:01:00", id: "10de:1c30", subsystem_id: "", iommu_group: 0}
|
||||
gpu-p2000-2: {node: pve3, path: "0000:02:00", id: "10de:1c30", subsystem_id: "", iommu_group: 0}
|
||||
gpu-p2000-3: {node: pve4, path: "0000:01:00", id: "10de:1c30", subsystem_id: "", iommu_group: 0}
|
||||
nodes: # Proxmox-Knoten (Cluster-Mitglieder), Ansible-Gruppen siehe platform/proxmox/ansible
|
||||
pve44: {ip: 192.168.100.210, gpu_host: true, zfs_tank: true}
|
||||
pve46: {ip: 192.168.100.209, gpu_host: true, zfs_tank: false} # eine Platte
|
||||
pve48: {ip: 192.168.100.212, gpu_host: true, zfs_tank: true}
|
||||
pve52: {ip: 192.168.100.207, gpu_host: true, zfs_tank: true}
|
||||
pve61: {ip: 192.168.100.214, gpu_host: false, zfs_tank: false} # Surface, Kiosk, keine Gäste
|
||||
|
||||
vms: # GPU-Rollen als VM (ADR-0001)
|
||||
gpu-main: {node: pve-ada, vmid: 201, cores: 6, memory_mb: 28672, disk_gb: 40, models_disk_gb: 100, gpu: gpu-ada-1, ip: 10.20.0.201/24, role: gpu-vllm, driver_generation: modern}
|
||||
gpu-flex: {node: pve-ada, vmid: 202, cores: 6, memory_mb: 28672, disk_gb: 40, models_disk_gb: 100, gpu: gpu-ada-2, ip: 10.20.0.202/24, role: gpu-vllm, driver_generation: modern}
|
||||
gpu-embed-1: {node: pve3, vmid: 221, cores: 2, memory_mb: 6144, disk_gb: 30, models_disk_gb: 20, gpu: gpu-p2000-1, ip: 10.20.0.221/24, role: gpu-llamacpp, profile: embed, driver_generation: legacy}
|
||||
gpu-embed-2: {node: pve4, vmid: 222, cores: 2, memory_mb: 6144, disk_gb: 30, models_disk_gb: 20, gpu: gpu-p2000-3, ip: 10.20.0.222/24, role: gpu-llamacpp, profile: embed, driver_generation: legacy}
|
||||
gpu-small: {node: pve3, vmid: 231, cores: 2, memory_mb: 8192, disk_gb: 30, models_disk_gb: 20, gpu: gpu-p2000-2, ip: 10.20.0.231/24, role: gpu-llamacpp, profile: chat, driver_generation: legacy}
|
||||
pci_mappings: # Datacenter → Resource Mappings → PCI; Pfad ohne Funktion = Grafik + Audio
|
||||
gpu-m4000-pve44: {node: pve44, path: "0000:01:00", id: "10de:13f1", subsystem_id: "", iommu_group: 1}
|
||||
gpu-p2000-pve44: {node: pve44, path: "0000:02:00", id: "10de:1c30", subsystem_id: "", iommu_group: 8}
|
||||
gpu-m4000-pve46: {node: pve46, path: "0000:01:00", id: "10de:13f1", subsystem_id: "", iommu_group: 1}
|
||||
gpu-p2000-pve46: {node: pve46, path: "0000:02:00", id: "10de:1c30", subsystem_id: "", iommu_group: 8}
|
||||
gpu-m4000-pve48: {node: pve48, path: "0000:01:00", id: "10de:13f1", subsystem_id: "", iommu_group: 1}
|
||||
gpu-p2000-pve48: {node: pve48, path: "0000:02:00", id: "10de:1c30", subsystem_id: "", iommu_group: 8}
|
||||
gpu-m4000-pve52: {node: pve52, path: "0000:01:00", id: "10de:13f1", subsystem_id: "", iommu_group: 1}
|
||||
|
||||
containers: # Infrastruktur als LXC (ADR-0001), HA + ZFS-Replikation (ADR-0002)
|
||||
infra: {node: pve1, vmid: 101, cores: 4, memory_mb: 12288, disk_gb: 60, ip: 10.20.0.101/24, ha: true, replicate_to: pve2}
|
||||
vms: # eine GPU-VM pro Knoten mit allen Karten (D2), 12 GB fest (RAM-Budget)
|
||||
gpu-pve44: {node: pve44, vmid: 244, cores: 4, memory_mb: 12288, disk_gb: 40, models_disk_gb: 60, gpus: [gpu-m4000-pve44, gpu-p2000-pve44], ip: 192.168.100.224/24, role: gpu-llamacpp, driver_generation: legacy}
|
||||
gpu-pve46: {node: pve46, vmid: 246, cores: 4, memory_mb: 12288, disk_gb: 40, models_disk_gb: 60, gpus: [gpu-m4000-pve46, gpu-p2000-pve46], ip: 192.168.100.226/24, role: gpu-llamacpp, driver_generation: legacy}
|
||||
gpu-pve48: {node: pve48, vmid: 248, cores: 4, memory_mb: 12288, disk_gb: 40, models_disk_gb: 60, gpus: [gpu-m4000-pve48, gpu-p2000-pve48], ip: 192.168.100.228/24, role: gpu-llamacpp, driver_generation: legacy}
|
||||
gpu-pve52: {node: pve52, vmid: 252, cores: 4, memory_mb: 12288, disk_gb: 40, models_disk_gb: 60, gpus: [gpu-m4000-pve52], ip: 192.168.100.232/24, role: gpu-llamacpp, driver_generation: legacy}
|
||||
|
||||
containers: # LXC (Container zuerst); HA + Replikation nur auf `tank`-Knoten (Task 2.2)
|
||||
ops: {node: pve46, vmid: 100, cores: 2, memory_mb: 1024, disk_gb: 16, ip: 192.168.100.220/24, storage: local-lvm, nesting: false} # Ansible/OpenTofu, kein HA (RAM-Budget)
|
||||
infra: {node: pve44, vmid: 101, cores: 4, memory_mb: 8192, disk_gb: 60, ip: 192.168.100.221/24, storage: tank, nesting: true, ha: true, replicate_to: [pve52, pve48]}
|
||||
|
||||
# Später, eigene Tasks: pbs (VM 110, pve52, Task 5.1, .222), grp-test (LXC, pve48 auf tank, Phase 8, .223).
|
||||
replication_schedule: "*/15"
|
||||
|
||||
Reference in New Issue
Block a user