claude-mpm 4.0.32__py3-none-any.whl → 4.1.0__py3-none-any.whl

claude_mpm/services/mcp_gateway/core/process_pool.py

@@ -0,0 +1,411 @@
+"""
+MCP Process Pool Manager
+========================
+
+Manages a pool of MCP server processes to prevent multiple instances
+and reduce startup overhead through connection reuse.
+
+WHY: MCP vector search servers load 400MB+ indexes on startup causing 11.9s delays.
+By maintaining a process pool and reusing connections, we eliminate this overhead.
+
+DESIGN DECISIONS:
+- Singleton process pool shared across all agent invocations
+- Pre-warm processes during framework initialization
+- Health checks and automatic restart of failed processes
+- Graceful shutdown and resource cleanup
+"""
+
+import asyncio
+import json
+import os
+import signal
+import subprocess
+import sys
+import threading
+import time
+from pathlib import Path
+from typing import Any, Dict, List, Optional, Tuple
+
+from claude_mpm.config.paths import paths
+from claude_mpm.core.logger import get_logger
+
+
+class MCPProcessPool:
+    """
+    Manages a pool of MCP server processes for efficient resource utilization.
+    
+    WHY: Prevent multiple MCP server instances from being spawned and
+    reduce startup overhead by reusing existing processes.
+    """
+    
+    _instance: Optional['MCPProcessPool'] = None
+    _lock = threading.Lock()
+    
+    def __new__(cls):
+        """Singleton pattern implementation."""
+        with cls._lock:
+            if cls._instance is None:
+                cls._instance = super().__new__(cls)
+                cls._instance._initialized = False
+            return cls._instance
+    
+    def __init__(self):
+        """Initialize the process pool manager."""
+        if self._initialized:
+            return
+            
+        self.logger = get_logger("MCPProcessPool")
+        self._initialized = True
+        
+        # Process tracking
+        self._processes: Dict[str, subprocess.Popen] = {}
+        self._process_info: Dict[str, Dict] = {}
+        self._startup_times: Dict[str, float] = {}
+        
+        # Configuration
+        self.max_processes = 3  # Maximum number of pooled processes
+        self.process_timeout = 300  # 5 minutes idle timeout
+        self.health_check_interval = 30  # Check process health every 30s
+        
+        # Paths
+        self.pool_dir = paths.claude_mpm_dir_hidden / "mcp" / "pool"
+        self.pool_dir.mkdir(parents=True, exist_ok=True)
+        
+        # Pre-warming flag
+        self._pre_warmed = False
+        
+        # Background health check task
+        self._health_check_task: Optional[asyncio.Task] = None
+        
+        # Setup cleanup handlers
+        self._setup_cleanup_handlers()
+        
+        self.logger.info("MCP Process Pool initialized")
+    
+    def _setup_cleanup_handlers(self):
+        """Setup signal handlers for cleanup on termination."""
+        def cleanup_handler(signum, frame):
+            self.logger.info(f"Received signal {signum}, cleaning up process pool")
+            self.cleanup_all()
+            
+        signal.signal(signal.SIGTERM, cleanup_handler)
+        signal.signal(signal.SIGINT, cleanup_handler)
+    
+    def get_or_create_process(self, server_name: str, config: Dict) -> Optional[subprocess.Popen]:
+        """
+        Get an existing process or create a new one for the given server.
+        
+        Args:
+            server_name: Name of the MCP server
+            config: Server configuration including command and args
+            
+        Returns:
+            Process handle or None if failed
+        """
+        start_time = time.time()
+        
+        # Check if we have a healthy existing process
+        if server_name in self._processes:
+            process = self._processes[server_name]
+            if self._is_process_healthy(process):
+                self.logger.info(f"Reusing existing process for {server_name} (PID: {process.pid})")
+                return process
+            else:
+                # Process is dead, clean it up
+                self.logger.warning(f"Process for {server_name} is dead, cleaning up")
+                self._cleanup_process(server_name)
+        
+        # Check if we've hit the process limit
+        if len(self._processes) >= self.max_processes:
+            # Find and clean up the oldest idle process
+            self._cleanup_oldest_idle_process()
+        
+        # Create new process
+        self.logger.info(f"Creating new process for {server_name}")
+        process = self._create_process(server_name, config)
+        
+        if process:
+            create_time = time.time() - start_time
+            self.logger.info(f"Process created for {server_name} in {create_time:.2f}s (PID: {process.pid})")
+            self._startup_times[server_name] = create_time
+        
+        return process
+    
+    def _create_process(self, server_name: str, config: Dict) -> Optional[subprocess.Popen]:
+        """
+        Create a new MCP server process.
+        
+        Args:
+            server_name: Name of the MCP server
+            config: Server configuration
+            
+        Returns:
+            Process handle or None if failed
+        """
+        try:
+            # Extract command and args from config
+            command = config.get("command", "")
+            args = config.get("args", [])
+            env = config.get("env", {})
+            cwd = config.get("cwd")
+            
+            # Build full command
+            full_command = [command] + args
+            
+            # Merge environment variables
+            process_env = os.environ.copy()
+            process_env.update(env)
+            
+            # Add timing instrumentation
+            process_env["MCP_STARTUP_TRACKING"] = "1"
+            process_env["MCP_SERVER_NAME"] = server_name
+            
+            # Start the process
+            process = subprocess.Popen(
+                full_command,
+                stdin=subprocess.PIPE,
+                stdout=subprocess.PIPE,
+                stderr=subprocess.PIPE,
+                env=process_env,
+                cwd=cwd,
+                bufsize=0  # Unbuffered for real-time communication
+            )
+            
+            # Store process info
+            self._processes[server_name] = process
+            self._process_info[server_name] = {
+                "pid": process.pid,
+                "started_at": time.time(),
+                "last_used": time.time(),
+                "config": config
+            }
+            
+            # Write process info to file for debugging
+            info_file = self.pool_dir / f"{server_name}_{process.pid}.json"
+            with open(info_file, 'w') as f:
+                json.dump(self._process_info[server_name], f, indent=2)
+            
+            return process
+            
+        except Exception as e:
+            self.logger.error(f"Failed to create process for {server_name}: {e}")
+            return None
+    
+    def _is_process_healthy(self, process: subprocess.Popen) -> bool:
+        """Check if a process is still running and healthy."""
+        if process.poll() is not None:
+            # Process has terminated
+            return False
+        
+        try:
+            # Send signal 0 to check if process is alive
+            os.kill(process.pid, 0)
+            return True
+        except (OSError, ProcessLookupError):
+            return False
+    
+    def _cleanup_process(self, server_name: str):
+        """Clean up a specific process."""
+        if server_name not in self._processes:
+            return
+        
+        process = self._processes[server_name]
+        
+        try:
+            # Try graceful shutdown first
+            if self._is_process_healthy(process):
+                process.terminate()
+                try:
+                    process.wait(timeout=5)
+                except subprocess.TimeoutExpired:
+                    # Force kill if graceful shutdown fails
+                    process.kill()
+                    process.wait()
+            
+            # Remove from tracking
+            del self._processes[server_name]
+            del self._process_info[server_name]
+            
+            # Clean up info file
+            for info_file in self.pool_dir.glob(f"{server_name}_*.json"):
+                info_file.unlink()
+            
+            self.logger.info(f"Cleaned up process for {server_name}")
+            
+        except Exception as e:
+            self.logger.warning(f"Error cleaning up process for {server_name}: {e}")
+    
+    def _cleanup_oldest_idle_process(self):
+        """Find and clean up the oldest idle process."""
+        if not self._process_info:
+            return
+        
+        # Find process with oldest last_used time
+        oldest_server = min(
+            self._process_info.keys(),
+            key=lambda k: self._process_info[k].get("last_used", 0)
+        )
+        
+        self.logger.info(f"Cleaning up oldest idle process: {oldest_server}")
+        self._cleanup_process(oldest_server)
+    
+    async def pre_warm_servers(self, configs: Dict[str, Dict]):
+        """
+        Pre-warm MCP servers during framework initialization.
+        
+        Args:
+            configs: Dictionary of server configurations
+        """
+        if self._pre_warmed:
+            self.logger.info("Servers already pre-warmed")
+            return
+        
+        self.logger.info(f"Pre-warming {len(configs)} MCP servers")
+        start_time = time.time()
+        
+        # Start all servers in parallel
+        tasks = []
+        for server_name, config in configs.items():
+            # Only pre-warm critical servers (like vector search)
+            if "vector" in server_name.lower() or config.get("pre_warm", False):
+                self.logger.info(f"Pre-warming {server_name}")
+                process = self.get_or_create_process(server_name, config)
+                if process:
+                    self.logger.info(f"Pre-warmed {server_name} (PID: {process.pid})")
+        
+        self._pre_warmed = True
+        total_time = time.time() - start_time
+        self.logger.info(f"Pre-warming completed in {total_time:.2f}s")
+    
+    async def start_health_monitoring(self):
+        """Start background health monitoring of processes."""
+        if self._health_check_task and not self._health_check_task.done():
+            return
+        
+        self._health_check_task = asyncio.create_task(self._health_check_loop())
+        self.logger.info("Started health monitoring")
+    
+    async def _health_check_loop(self):
+        """Background loop to check process health."""
+        while True:
+            try:
+                await asyncio.sleep(self.health_check_interval)
+                
+                # Check each process
+                dead_processes = []
+                for server_name, process in self._processes.items():
+                    if not self._is_process_healthy(process):
+                        dead_processes.append(server_name)
+                
+                # Clean up dead processes
+                for server_name in dead_processes:
+                    self.logger.warning(f"Process {server_name} is dead, cleaning up")
+                    self._cleanup_process(server_name)
+                
+                # Check for idle timeout
+                current_time = time.time()
+                idle_processes = []
+                for server_name, info in self._process_info.items():
+                    last_used = info.get("last_used", current_time)
+                    if current_time - last_used > self.process_timeout:
+                        idle_processes.append(server_name)
+                
+                # Clean up idle processes
+                for server_name in idle_processes:
+                    self.logger.info(f"Process {server_name} idle timeout, cleaning up")
+                    self._cleanup_process(server_name)
+                    
+            except Exception as e:
+                self.logger.error(f"Error in health check loop: {e}")
+    
+    def mark_process_used(self, server_name: str):
+        """Mark a process as recently used."""
+        if server_name in self._process_info:
+            self._process_info[server_name]["last_used"] = time.time()
+    
+    def get_startup_metrics(self) -> Dict[str, float]:
+        """Get startup time metrics for all servers."""
+        return self._startup_times.copy()
+    
+    def get_pool_status(self) -> Dict[str, Any]:
+        """Get current status of the process pool."""
+        return {
+            "active_processes": len(self._processes),
+            "max_processes": self.max_processes,
+            "pre_warmed": self._pre_warmed,
+            "processes": {
+                name: {
+                    "pid": info.get("pid"),
+                    "uptime": time.time() - info.get("started_at", time.time()),
+                    "idle_time": time.time() - info.get("last_used", time.time())
+                }
+                for name, info in self._process_info.items()
+            },
+            "startup_metrics": self._startup_times
+        }
+    
+    def cleanup_all(self):
+        """Clean up all processes in the pool."""
+        self.logger.info("Cleaning up all processes in pool")
+        
+        # Stop health monitoring
+        if self._health_check_task:
+            self._health_check_task.cancel()
+        
+        # Clean up all processes
+        for server_name in list(self._processes.keys()):
+            self._cleanup_process(server_name)
+        
+        self.logger.info("Process pool cleanup completed")
+
+
+# Global instance
+_pool: Optional[MCPProcessPool] = None
+
+
+def get_process_pool() -> MCPProcessPool:
+    """Get the global MCP process pool instance."""
+    global _pool
+    if _pool is None:
+        _pool = MCPProcessPool()
+    return _pool
+
+
+async def pre_warm_mcp_servers():
+    """Pre-warm MCP servers from configuration."""
+    pool = get_process_pool()
+    
+    # Load MCP configurations
+    configs = {}
+    
+    # Check .claude.json for MCP server configs
+    claude_config_path = Path.home() / ".claude.json"
+    if not claude_config_path.exists():
+        # Try project-local config
+        claude_config_path = Path.cwd() / ".claude.json"
+    
+    if claude_config_path.exists():
+        try:
+            with open(claude_config_path, 'r') as f:
+                config_data = json.load(f)
+                mcp_servers = config_data.get("mcpServers", {})
+                configs.update(mcp_servers)
+        except Exception as e:
+            get_logger("MCPProcessPool").warning(f"Failed to load Claude config: {e}")
+    
+    # Check .mcp.json for additional configs
+    mcp_config_path = Path.cwd() / ".mcp.json"
+    if mcp_config_path.exists():
+        try:
+            with open(mcp_config_path, 'r') as f:
+                config_data = json.load(f)
+                mcp_servers = config_data.get("mcpServers", {})
+                configs.update(mcp_servers)
+        except Exception as e:
+            get_logger("MCPProcessPool").warning(f"Failed to load MCP config: {e}")
+    
+    if configs:
+        await pool.pre_warm_servers(configs)
+        await pool.start_health_monitoring()
+    
+    return pool

claude_mpm/services/mcp_gateway/server/stdio_server.py

@@ -15,7 +15,10 @@ use JSON-RPC protocol, and exit cleanly when stdin closes.
 import asyncio
 import json
 import logging
+import os
 import sys
+import time
+from pathlib import Path
 from typing import Any, Dict, Optional
 
 # Import MCP SDK components
@@ -28,6 +31,7 @@ from mcp.types import TextContent, Tool
 from pydantic import BaseModel
 
 from claude_mpm.core.logger import get_logger
+from claude_mpm.services.mcp_gateway.core.singleton_manager import get_gateway_manager
 
 # Import unified ticket tool if available
 try:
@@ -135,6 +139,11 @@ class SimpleMCPServer:
         self.name = name
         self.version = version
         self.logger = get_logger("MCPStdioServer")
+        self.startup_time = time.time()
+        
+        # Log startup timing
+        self.logger.info(f"Initializing MCP server {name} v{version}")
+        start_time = time.time()
 
         # Apply backward compatibility patches before creating server
         apply_backward_compatibility_patches()
@@ -144,6 +153,10 @@ class SimpleMCPServer:
 
         # Register default tools
         self._register_tools()
+        
+        # Log initialization time
+        init_time = time.time() - start_time
+        self.logger.info(f"MCP server initialized in {init_time:.2f} seconds")
 
     async def _summarize_content(
         self, content: str, style: str, max_length: int