Hyungi Ahn
d47c04317c
scheduler_status, queue_status, run_verify 추가. MCP 10개 도구 + NanoClaude wrapper + pre-route 키워드. worker.py trace print 제거. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
74 lines
2.4 KiB
Python
74 lines
2.4 KiB
Python
"""Run predefined verify commands from infra_inventory.md."""
|
|
|
|
from __future__ import annotations
|
|
|
|
from datetime import datetime, timezone
|
|
|
|
from ..config import HOSTS
|
|
from .ssh import run_command, run_local, SSHError, _is_local_host
|
|
|
|
|
|
def _now() -> str:
|
|
return datetime.now(timezone.utc).isoformat()
|
|
|
|
|
|
# Predefined verify commands (from infra_inventory.md)
|
|
VERIFY_COMMANDS = {
|
|
"gpu-snapshot": {
|
|
"host": "gpu",
|
|
"cmd": "docker compose ls 2>/dev/null; echo '---'; ollama list 2>/dev/null; echo '---'; pgrep -af 'Plex Media' | head -1 2>/dev/null",
|
|
"desc": "GPU 서버 전체 스냅샷 (docker, ollama, plex)",
|
|
},
|
|
"macmini-snapshot": {
|
|
"host": "macmini",
|
|
"cmd": "bash -lc 'launchctl list | grep com.user; curl -s localhost:8800/v1/models 2>/dev/null | python3 -m json.tool 2>/dev/null; ollama list 2>/dev/null'",
|
|
"desc": "Mac mini 전체 스냅샷 (launchd, MLX, ollama)",
|
|
},
|
|
"docserver-health": {
|
|
"host": "gpu",
|
|
"cmd": "curl -sf http://localhost:8000/health",
|
|
"desc": "Document Server 헬스체크",
|
|
},
|
|
"config-model-match": {
|
|
"host": "gpu",
|
|
"cmd": "grep -E 'model:' ~/Documents/code/hyungi_Document_Server/config.yaml 2>/dev/null; echo '---'; ollama list 2>/dev/null",
|
|
"desc": "config.yaml 모델과 실제 설치 모델 비교",
|
|
},
|
|
}
|
|
|
|
|
|
async def run_verify(check_name: str) -> dict:
|
|
"""Run a predefined verify command."""
|
|
if check_name not in VERIFY_COMMANDS:
|
|
available = ", ".join(VERIFY_COMMANDS.keys())
|
|
return {
|
|
"ok": False, "checked_at": _now(),
|
|
"error_type": "parse_error",
|
|
"error": f"알 수 없는 체크: '{check_name}'. 사용 가능: {available}",
|
|
"data": [], "summary": "", "raw": "",
|
|
}
|
|
|
|
spec = VERIFY_COMMANDS[check_name]
|
|
cfg = HOSTS[spec["host"]]
|
|
|
|
try:
|
|
if _is_local_host(cfg):
|
|
stdout, _ = await run_local(spec["cmd"], timeout=15)
|
|
else:
|
|
stdout, _ = await run_command(cfg, spec["cmd"], timeout=15)
|
|
except SSHError as e:
|
|
return {
|
|
"ok": False, "checked_at": _now(),
|
|
"error_type": e.error_type, "error": str(e),
|
|
"data": [], "summary": spec["desc"], "raw": "",
|
|
}
|
|
|
|
return {
|
|
"ok": True,
|
|
"checked_at": _now(),
|
|
"data": stdout.strip().splitlines(),
|
|
"summary": spec["desc"],
|
|
"error": "",
|
|
"raw": stdout.strip(),
|
|
}
|