- effi-stt: Déchargement Whisper après 300s inactivité (INACTIVITY_TIMEOUT) * Nouveaux endpoints: /unload, /reload, /stats * Modelo rechargé à la première requête - ComfyUI: Script monitor qui décharge VRAM après inactivité * Vérifie usage VRAM toutes les 60s * Exécute workflow UnloadAllModelsNode après 300s inactivité - Kreuzberg: Manager qui redémarre le container après seuil mémoire + inactivité * Restart si > 2GB pendant > 10min inactif - Documentation: docs/MEMOIRE-GESTION.md
175 lines
6.2 KiB
Python
Executable File
175 lines
6.2 KiB
Python
Executable File
#!/usr/bin/env python3
|
|
# Kreuzberg Memory Manager - Redémarre le container après inactivité
|
|
# Usage: python3 kreuzberg-memory-manager.py [--interval 60] [--timeout 600] [--max-memory 2]
|
|
|
|
import requests
|
|
import time
|
|
import argparse
|
|
import logging
|
|
import subprocess
|
|
import sys
|
|
|
|
logging.basicConfig(
|
|
level=logging.INFO,
|
|
format='%(asctime)s - %(levelname)s - %(message)s'
|
|
)
|
|
logger = logging.getLogger(__name__)
|
|
|
|
DEFAULT_INTERVAL = 60
|
|
DEFAULT_TIMEOUT = 600
|
|
DEFAULT_MAX_MEMORY = 2.0
|
|
KREUZBERG_API = "http://localhost:8300"
|
|
CONTAINER_NAME = "kreuzberg"
|
|
|
|
class KreuzbergMemoryManager:
|
|
def __init__(self, interval=DEFAULT_INTERVAL, timeout=DEFAULT_TIMEOUT, max_memory_gb=DEFAULT_MAX_MEMORY):
|
|
self.interval = interval
|
|
self.timeout = timeout
|
|
self.max_memory_gb = max_memory_gb
|
|
self.last_activity = time.time()
|
|
self.running = True
|
|
self.container_id = None
|
|
|
|
import signal
|
|
signal.signal(signal.SIGINT, self.signal_handler)
|
|
signal.signal(signal.SIGTERM, self.signal_handler)
|
|
|
|
self.find_container()
|
|
|
|
def signal_handler(self, signum, frame):
|
|
logger.info("Signal reçu, arrêt du manager...")
|
|
self.running = False
|
|
|
|
def find_container(self):
|
|
try:
|
|
result = subprocess.run(
|
|
["docker", "ps", "-q", "--filter", f"name={CONTAINER_NAME}"],
|
|
capture_output=True, text=True
|
|
)
|
|
if result.stdout.strip():
|
|
self.container_id = result.stdout.strip()
|
|
logger.info(f"Container trouvé: {self.container_id[:12]}")
|
|
else:
|
|
logger.warning(f"Container {CONTAINER_NAME} non trouvé")
|
|
except Exception as e:
|
|
logger.error(f"Erreur recherche container: {e}")
|
|
|
|
def get_container_memory(self):
|
|
if not self.container_id:
|
|
return None
|
|
try:
|
|
result = subprocess.run(
|
|
["docker", "stats", "--no-stream", "--format", "{{.MemUsage}}", self.container_id],
|
|
capture_output=True, text=True
|
|
)
|
|
if result.stdout.strip():
|
|
mem_str = result.stdout.strip()
|
|
return self.parse_memory(mem_str)
|
|
except Exception as e:
|
|
logger.debug(f"Erreur stats: {e}")
|
|
return None
|
|
|
|
def parse_memory(self, mem_str):
|
|
try:
|
|
parts = mem_str.split("/")
|
|
if len(parts) >= 1:
|
|
used = parts[0].strip()
|
|
if "GiB" in used:
|
|
return float(used.replace("GiB", "")) * 1024
|
|
elif "MiB" in used:
|
|
return float(used.replace("MiB", ""))
|
|
elif "GB" in used:
|
|
return float(used.replace("GB", "")) * 1024
|
|
elif "MB" in used:
|
|
return float(used.replace("MB", ""))
|
|
except:
|
|
pass
|
|
return None
|
|
|
|
def check_api_activity(self):
|
|
try:
|
|
response = requests.get(f"{KREUZBERG_API}/health", timeout=5)
|
|
if response.status_code in [200, 404]:
|
|
self.last_activity = time.time()
|
|
return True
|
|
except:
|
|
pass
|
|
return False
|
|
|
|
def restart_container(self):
|
|
logger.info(f"🔄 Redémarrage du container {CONTAINER_NAME}...")
|
|
try:
|
|
subprocess.run(["docker", "restart", CONTAINER_NAME], check=True)
|
|
logger.info("✅ Container redémarré")
|
|
time.sleep(5)
|
|
self.find_container()
|
|
return True
|
|
except Exception as e:
|
|
logger.error(f"❌ Erreur redémarrage: {e}")
|
|
return False
|
|
|
|
def run(self):
|
|
logger.info(f"🚀 Démarrage Kreuzberg Memory Manager")
|
|
logger.info(f" Intervalle: {self.interval}s")
|
|
logger.info(f" Timeout inactivité: {self.timeout}s")
|
|
logger.info(f" Seuil mémoire: {self.max_memory_gb}GB")
|
|
|
|
idle_time = 0
|
|
|
|
while self.running:
|
|
try:
|
|
memory_mb = self.get_container_memory()
|
|
|
|
if memory_mb is not None:
|
|
memory_gb = memory_mb / 1024
|
|
|
|
is_active = self.check_api_activity()
|
|
|
|
if is_active:
|
|
idle_time = 0
|
|
self.last_activity = time.time()
|
|
logger.debug(f"📊 Mémoire: {memory_gb:.2f}GB, API active")
|
|
else:
|
|
idle_time += self.interval
|
|
logger.debug(f"📊 Mémoire: {memory_gb:.2f}GB, Inactif: {idle_time}s")
|
|
|
|
if memory_gb > self.max_memory_gb:
|
|
logger.warning(f"⚠️ Mémoire ({memory_gb:.2f}GB) > seuil ({self.max_memory_gb}GB)")
|
|
if idle_time >= self.timeout:
|
|
logger.info(f"⏰ Seuil timeout atteint ({idle_time}s >= {self.timeout}s)")
|
|
self.restart_container()
|
|
idle_time = 0
|
|
|
|
time.sleep(self.interval)
|
|
|
|
except Exception as e:
|
|
logger.error(f"Erreur: {e}")
|
|
time.sleep(self.interval)
|
|
|
|
logger.info("👋 Manager arrêté")
|
|
|
|
def main():
|
|
parser = argparse.ArgumentParser(description="Kreuzberg Memory Manager")
|
|
parser.add_argument("--interval", type=int, default=DEFAULT_INTERVAL,
|
|
help=f"Intervalle (défaut: {DEFAULT_INTERVAL}s)")
|
|
parser.add_argument("--timeout", type=int, default=DEFAULT_TIMEOUT,
|
|
help=f"Timeout inactivité (défaut: {DEFAULT_TIMEOUT}s)")
|
|
parser.add_argument("--max-memory", type=float, default=DEFAULT_MAX_MEMORY,
|
|
help=f"Seuil mémoire GB (défaut: {DEFAULT_MAX_MEMORY}GB)")
|
|
parser.add_argument("--debug", action="store_true", help="Mode debug")
|
|
|
|
args = parser.parse_args()
|
|
|
|
if args.debug:
|
|
logging.getLogger().setLevel(logging.DEBUG)
|
|
|
|
manager = KreuzbergMemoryManager(
|
|
interval=args.interval,
|
|
timeout=args.timeout,
|
|
max_memory_gb=args.max_memory
|
|
)
|
|
manager.run()
|
|
|
|
if __name__ == "__main__":
|
|
main()
|