Files
runpod/model-orchestrator/orchestrator_subprocess.py

324 lines
10 KiB
Python

#!/usr/bin/env python3
"""
AI Model Orchestrator for RunPod (Process-Based)
Manages sequential loading of AI models using subprocess instead of Docker
Simplified architecture for RunPod's containerized environment:
- No Docker-in-Docker complexity
- Direct process management via subprocess
- Models run as Python background processes
- GPU memory efficient (sequential model loading)
"""
import asyncio
import logging
import os
import subprocess
import time
import signal
from typing import Dict, Optional, Any
from pathlib import Path
import httpx
import yaml
import psutil
from fastapi import FastAPI, Request, HTTPException
from fastapi.responses import JSONResponse, StreamingResponse
from pydantic import BaseModel
# Configure logging
logging.basicConfig(
level=logging.INFO,
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
)
logger = logging.getLogger(__name__)
# FastAPI app
app = FastAPI(title="AI Model Orchestrator (Process-Based)", version="2.0.0")
# Global state
current_model: Optional[str] = None
model_processes: Dict[str, subprocess.Popen] = {}
model_registry: Dict[str, Dict[str, Any]] = {}
config: Dict[str, Any] = {}
def load_model_registry():
"""Load model registry from models.yaml"""
global model_registry, config
config_path = os.getenv("MODELS_CONFIG", "/workspace/ai/model-orchestrator/models.yaml")
logger.info(f"Loading model registry from {config_path}")
with open(config_path, 'r') as f:
data = yaml.safe_load(f)
model_registry = data.get('models', {})
config = data.get('config', {})
logger.info(f"Loaded {len(model_registry)} models")
for model_name, model_config in model_registry.items():
logger.info(f" - {model_name}: {model_config.get('type')} ({model_config.get('framework')})")
async def start_model_process(model_name: str) -> bool:
"""Start a model as a subprocess"""
global current_model, model_processes
if model_name not in model_registry:
logger.error(f"Model {model_name} not found in registry")
return False
model_config = model_registry[model_name]
# Stop current model if running
if current_model and current_model != model_name:
await stop_model_process(current_model)
# Check if already running
if model_name in model_processes:
proc = model_processes[model_name]
if proc.poll() is None: # Still running
logger.info(f"Model {model_name} already running")
return True
logger.info(f"Starting model {model_name}...")
try:
# Get service command from config
service_script = model_config.get('service_script')
if not service_script:
logger.error(f"No service_script defined for {model_name}")
return False
script_path = Path(f"/workspace/ai/{service_script}")
if not script_path.exists():
logger.error(f"Service script not found: {script_path}")
return False
# Start process
port = model_config.get('port', 8000)
env = os.environ.copy()
env.update({
'HF_TOKEN': os.getenv('HF_TOKEN', ''),
'PORT': str(port),
'HOST': '0.0.0.0',
'MODEL_NAME': model_config.get('model_name', model_name)
})
# Use venv python if it exists
script_dir = script_path.parent
venv_python = script_dir / 'venv' / 'bin' / 'python3'
python_cmd = str(venv_python) if venv_python.exists() else 'python3'
proc = subprocess.Popen(
[python_cmd, str(script_path)],
env=env,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
preexec_fn=os.setsid # Create new process group
)
model_processes[model_name] = proc
# Wait for service to be ready
max_wait = model_config.get('startup_time_seconds', 120)
start_time = time.time()
while time.time() - start_time < max_wait:
if proc.poll() is not None:
logger.error(f"Process for {model_name} exited prematurely")
return False
try:
async with httpx.AsyncClient() as client:
response = await client.get(
f"http://localhost:{port}/health",
timeout=5.0
)
if response.status_code == 200:
logger.info(f"Model {model_name} is ready on port {port}")
current_model = model_name
return True
except:
await asyncio.sleep(2)
logger.error(f"Model {model_name} failed to start within {max_wait}s")
await stop_model_process(model_name)
return False
except Exception as e:
logger.error(f"Error starting {model_name}: {e}")
return False
async def stop_model_process(model_name: str):
"""Stop a running model process"""
global model_processes, current_model
if model_name not in model_processes:
logger.warning(f"Model {model_name} not in process registry")
return
proc = model_processes[model_name]
if proc.poll() is None: # Still running
logger.info(f"Stopping model {model_name}...")
try:
# Send SIGTERM to process group
os.killpg(os.getpgid(proc.pid), signal.SIGTERM)
# Wait for graceful shutdown
try:
proc.wait(timeout=10)
except subprocess.TimeoutExpired:
# Force kill if not terminated
os.killpg(os.getpgid(proc.pid), signal.SIGKILL)
proc.wait()
logger.info(f"Model {model_name} stopped")
except Exception as e:
logger.error(f"Error stopping {model_name}: {e}")
del model_processes[model_name]
if current_model == model_name:
current_model = None
def get_model_for_endpoint(endpoint: str) -> Optional[str]:
"""Determine which model handles this endpoint"""
for model_name, model_config in model_registry.items():
if endpoint.startswith(model_config.get('endpoint', '')):
return model_name
return None
@app.on_event("startup")
async def startup_event():
"""Initialize on startup"""
logger.info("Starting AI Model Orchestrator (Process-Based)")
load_model_registry()
@app.on_event("shutdown")
async def shutdown_event():
"""Cleanup on shutdown"""
logger.info("Shutting down orchestrator...")
for model_name in list(model_processes.keys()):
await stop_model_process(model_name)
@app.get("/health")
async def health_check():
"""Health check endpoint"""
return {
"status": "healthy",
"current_model": current_model,
"active_processes": len(model_processes),
"available_models": list(model_registry.keys())
}
@app.get("/v1/models")
async def list_models_openai():
"""OpenAI-compatible models listing endpoint"""
models_list = []
for model_name, model_info in model_registry.items():
models_list.append({
"id": model_name,
"object": "model",
"created": int(time.time()),
"owned_by": "pivoine-gpu",
"permission": [],
"root": model_name,
"parent": None,
})
return {
"object": "list",
"data": models_list
}
@app.api_route("/{path:path}", methods=["GET", "POST", "PUT", "DELETE", "PATCH"])
async def proxy_request(request: Request, path: str):
"""Proxy requests to appropriate model service"""
endpoint = f"/{path}"
# Determine which model should handle this
target_model = get_model_for_endpoint(endpoint)
if not target_model:
raise HTTPException(status_code=404, detail=f"No model configured for endpoint: {endpoint}")
# Ensure model is running
if current_model != target_model:
logger.info(f"Switching to model {target_model}")
success = await start_model_process(target_model)
if not success:
raise HTTPException(status_code=503, detail=f"Failed to start model {target_model}")
# Proxy the request
model_config = model_registry[target_model]
target_url = f"http://localhost:{model_config['port']}/{path}"
# Get request details
method = request.method
headers = dict(request.headers)
headers.pop('host', None) # Remove host header
body = await request.body()
# Check if this is a streaming request
is_streaming = False
if method == "POST" and body:
try:
import json
body_json = json.loads(body)
is_streaming = body_json.get('stream', False)
except:
pass
logger.info(f"Proxying {method} request to {target_url} (streaming: {is_streaming})")
try:
if is_streaming:
# For streaming requests, use httpx streaming and yield chunks
async def stream_response():
async with httpx.AsyncClient(timeout=300.0) as client:
async with client.stream(method, target_url, content=body, headers=headers) as response:
async for chunk in response.aiter_bytes():
yield chunk
return StreamingResponse(
stream_response(),
media_type="text/event-stream",
headers={"Cache-Control": "no-cache", "Connection": "keep-alive"}
)
else:
# For non-streaming requests, use the original behavior
async with httpx.AsyncClient(timeout=300.0) as client:
response = await client.request(
method=method,
url=target_url,
headers=headers,
content=body
)
return JSONResponse(
content=response.json() if response.headers.get('content-type') == 'application/json' else response.text,
status_code=response.status_code
)
except Exception as e:
logger.error(f"Error proxying request: {e}")
raise HTTPException(status_code=502, detail=str(e))
if __name__ == "__main__":
import uvicorn
port = int(os.getenv("PORT", "9000"))
host = os.getenv("HOST", "0.0.0.0")
logger.info(f"Starting orchestrator on {host}:{port}")
uvicorn.run(app, host=host, port=port, log_level="info")