Files
infrastructure/llm-server/llm/ansible
spektr 026c840029 LLM-Server: M2-Vertical-Slice in Betrieb - Swift-Stack läuft, Inferenz nachgewiesen
Erster produktiver Start des llama.cpp-Testing-Stacks auf ghostnet-ki
(31.10.2026). Fix-Liste aus den Live-Läufen:
- Quadlet-Exec ohne Binärnamen (AMD-Image hat ENTRYPOINT /app/llama-server;
  GNAI-010 v0.6 dokumentiert den Befund)
- Deterministischer daemon-reload vor Unit-Zugriff (Handler allein griff nicht,
  wenn ein vorheriger Lauf die Quadlets gerendert hatte)
- Netz-Abhängigkeit Wants/After im Container-Quadlet (Netzwerk-Unit wurde
  nie mitgestartet -> network not found)
- Stopp-Bedingung für Profilwechsel korrigiert (nur aktive Units stoppen)
- Health-Tool-Kette curl/wget/bash-/dev/tcp mit Pfad-Normalisierung; curl -f
  als Gesundheitssignal (Statuszeile stand nie im stdout)
- chdir für alle runuser-Aufrufe der Health-Checks

Nachweise: Health 200 OK, /v1/models Alias korrekt, Idempotenz changed=0,
VRAM 12,6/13,0 GiB von je 16 GiB auf beiden Karten (Layer-Split 1:1),
echte Chat-Completion mit System-Fingerprint des gepinnten Builds,
Reasoning getrennt von der Antwort.

Generated by Mistral Vibe.
Co-Authored-By: Mistral Vibe <vibe@mistral.ai>
2026-11-01 17:32:17 +01:00
..

LLM serving stack – Ansible control project

Executes the LLM serving stack for exactly one environment on ghostnet-ki (GNAI-008 section 7.4). Host baseline, Podman runtime, workload accounts and AMD GPU access are managed by ../../host/ansible; this project never changes host state.

Layout

  • inventories/<environment>/ – one inventory per KI environment (currently testing); deployment uses the ansible-deploy account with become: true.
  • playbooks/ai.yml – applies model_store then inference_engine.
  • roles/model_store/ – shared, pinned model weights under /var/lib/ghostnet-ai/shared/models (GNAI-010 sections 3 and 4).
  • roles/inference_engine/ – rootless Quadlets, network, image provisioning and the exclusive profile switch for llama.cpp (GNAI-010 sections 5 and 6).

Control node setup

Use the repository-wide control venv built from ../../../shared/ansible/requirements-control.txt (ansible-core==2.21.4), for example:

python3 -m venv .venv
.venv/bin/pip install -r shared/ansible/requirements-control.txt
cd llm-server/llm/ansible
../../../.venv/bin/ansible-playbook -i inventories/testing playbooks/ai.yml --syntax-check

Usage (Testing)

All commands use the operator deploy key ~/.ssh/id_ed25519_ghostnet_ai_ansible_deploy (see AGENTS.md; no ssh-agent required):

# Preview only – no changes, no downloads, no service start:
../../../.venv/bin/ansible-playbook -i inventories/testing playbooks/ai.yml \
  --check --diff --private-key ~/.ssh/id_ed25519_ghostnet_ai_ansible_deploy

# Provision model weights only (tag models):
../../../.venv/bin/ansible-playbook -i inventories/testing playbooks/ai.yml \
  --tags models --private-key ~/.ssh/id_ed25519_ghostnet_ai_ansible_deploy

# Render quadlets and start/verify the selected profile (tag service):
../../../.venv/bin/ansible-playbook -i inventories/testing playbooks/ai.yml \
  --tags service --private-key ~/.ssh/id_ed25519_ghostnet_ai_ansible_deploy

inference_engine_active_profile selects which rendered quadlet is started; all other profiles stay rendered but stopped. Check mode downloads no model and starts no service. A second identical run must report changed=0 (GNAI-010 section 10.4).

The full technical acceptance (64k/128k start, VRAM distribution, prefill ladder, restart behaviour) is an operator-run protocol documented in GNAI-010; this role automates deployment and basic health only.