Refactor llama_server role for Windows support and EL-specific setup

- Added Windows-specific tasks for llama_server role, including installation of dependencies, setup of services using NSSM, and configuration of firewall rules.
- Introduced a new task file for Windows Exporter setup.
- Consolidated EL-specific setup tasks into a separate include file.
- Updated inventory to include additional host for llama server.
- Adjusted error handling in Prometheus service removal task to prevent failures.
This commit is contained in:
2026-06-01 14:59:30 +02:00
parent bb8b14dcca
commit 3775fa4b6e
12 changed files with 1306 additions and 355 deletions
+29 -1
View File
@@ -9,8 +9,10 @@ llama_server_tag: "server-vulkan"
llama_server_extra_groups: "users,video"
llama_server_devices: []
llama_server_env: {}
llama_server_port: 8080
llama_server_port: 11434
llama_server_preset_global: {}
llama_server_host: "127.0.0.1"
llama_server_argv_extra: []
# Llama Exporter configuration
llama_server_exporter_enabled: true
@@ -24,3 +26,29 @@ llama_server_exporter_scrape_timeout: 5
# Llama.cpp scrape targets
llama_server_exporter_targets: []
# Windows-specific defaults
llama_server_windows_install_dir: "C:\\llama_server"
llama_server_windows_user: "llama"
llama_server_windows_working_dir: "C:\\llama_server"
llama_server_windows_version: "b9442"
llama_server_windows_backend: "cuda"
llama_server_cuda_version: "13.3"
llama_server_windows_nssm_dir: "C:\\nssm"
llama_server_windows_nssm_version: "2.24"
llama_server_windows_nssm_url: "https://nssm.cc/release/nssm-{{ llama_server_windows_nssm_version }}.zip"
# Llama-server binary defaults for Windows
llama_server_windows_bin_name: "llama-server.exe"
# Windows Python defaults
llama_server_windows_python_version: "3.12.10"
llama_server_windows_python_path: "C:\\Program Files\\Python{{ llama_server_windows_python_version | regex_replace('^(\\d+)\\.(\\d+)\\..*$', '\\1\\2') }}\\python.exe"
llama_server_windows_vcredist_url: "https://aka.ms/vs/17/release/vc_redist.x64.exe"
# Windows Exporter defaults
llama_server_windows_exporter_enabled: true
llama_server_windows_exporter_port: 9550
llama_server_windows_exporter_interval: 15
llama_server_windows_exporter_timeout: 5
llama_server_windows_exporter_targets: []