Files
spektr 148104de4e gnai-ai-backup: Boot-Rennen um den NFS-Mount beheben (RequiresMountsFor)
Befund 09.11.2026: Beim Host-Neustart lief gnai-ai-backup.service um
05:05:45 mit Exit 21, weil /var/lib/ghostnet-pipeline noch nicht gemountet
war. Ursache: Die Service-Unit enthielt After=var-lib-ghostnet-pipeline.mount
- systemd escapt das '-' als \x2d, die reale Mount-Unit heisst
var-lib-ghostnet\x2dpipeline.mount; der literale Name existierte nie
(not-found) und ein blosses After= zieht die Mount-Unit ohne Requires auch
nicht in die Timer-Start-Transaktion.

Fix: RequiresMountsFor={{ ai_backup_nfs_mountpoint }} in der Service-Unit
(ergibt Requires=+After= auf die korrekt escapte Mount-Unit; systemd leitet
den Unit-Namen aus dem Pfad ab). Bei NFS-Ausfall nach Boot schlaegt jetzt die
Mount-Unit fehl und der Service mit Dependency-Fehler statt stillem Exit 21.

Verifiziert 09.11.2026 (Deploy via playbooks/backup.yml, production-Inventar,
ansible-deploy): systemctl show/list-dependencies listen die Mount-Unit unter
Requires=/After=; manueller systemctl start -> Result=success, Archiv
ghostnet-ai-backup-20261109-0544.tar.gz auf dem NFS-Share mit sha256-Sidecar
(sha256sum -c OK, 123 Members, keine *.gguf); Rollen-Zweitlauf changed=0.
Doku: ai_backup-README, GNAI-008 14.6 (Befund als geloest markiert) + Changelog
0.21, GNAI-000 KI-Host-Zeile aktualisiert.

Generated by Mistral Vibe.
Co-Authored-By: Mistral Vibe <vibe@mistral.ai>
2026-11-09 05:45:55 +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.