feat: 添加监控路由和相关功能,更新依赖项以支持新特性,增强服务监控能力

This commit is contained in:
iHeyTang
2025-08-14 14:40:42 +08:00
parent 6240b98c5b
commit 9c72bffd21
5 changed files with 456 additions and 2 deletions

View File

@@ -6,3 +6,4 @@ aiohttp
boto3
aiosqlite
sqlalchemy
psutil

View File

@@ -9,6 +9,7 @@ from pathlib import Path
# 加载 .env 文件
try:
from dotenv import load_dotenv
env_path = Path(__file__).parent / ".env"
load_dotenv(env_path)
print(f"已加载环境变量文件: {env_path}")
@@ -25,5 +26,10 @@ sys.path.insert(0, project_root)
from workflow_service.main import web_app
import uvicorn
print("🚀 启动 ComfyUI 工作流服务监控...")
print("📊 监控页面地址: http://localhost:18000/monitor")
print("🔍 健康检查: http://localhost:18000/monitor/health")
print("📈 API 文档: http://localhost:18000/docs")
if __name__ == "__main__":
uvicorn.run(web_app, host="0.0.0.0", port=18000)

View File

@@ -254,3 +254,21 @@ async def get_running_workflow_runs() -> List[dict]:
runs = result.scalars().all()
return [run.to_dict() for run in runs]
async def get_workflow_runs_recent(start_time: datetime, end_time: datetime) -> List[dict]:
"""获取指定时间范围内的最近工作流运行记录"""
async with AsyncSessionLocal() as session:
stmt = (
select(WorkflowRun)
.where(
WorkflowRun.created_at >= start_time,
WorkflowRun.created_at <= end_time
)
.order_by(WorkflowRun.created_at.desc())
)
result = await session.execute(stmt)
runs = result.scalars().all()
return [run.to_dict() for run in runs]

View File

@@ -6,7 +6,7 @@ from fastapi.middleware.cors import CORSMiddleware
from workflow_service.database.api import init_db
from workflow_service.config import Settings
from workflow_service.routes import service, workflow, run, runx
from workflow_service.routes import service, workflow, run, runx, monitor
settings = Settings()
@@ -25,6 +25,7 @@ web_app.include_router(service.service_router)
web_app.include_router(workflow.workflow_router)
web_app.include_router(run.run_router)
web_app.include_router(runx.runx_router)
web_app.include_router(monitor.monitor_router)
@web_app.on_event("startup")

View File

@@ -0,0 +1,428 @@
from fastapi import APIRouter, HTTPException
from fastapi.responses import HTMLResponse
import psutil
import asyncio
from datetime import datetime, timedelta
from typing import Dict, Any, List
import json
from workflow_service.config import settings
from workflow_service.database.api import get_workflow_runs_recent, get_workflow_run_nodes
from workflow_service.comfy.comfy_queue import WorkflowQueueManager
monitor_router = APIRouter(prefix="/monitor", tags=["监控"])
# 全局队列管理器实例
queue_manager = WorkflowQueueManager()
@monitor_router.get("/", response_class=HTMLResponse)
async def monitor_dashboard():
"""监控仪表板页面"""
html_content = """
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>ComfyUI 工作流服务监控</title>
<style>
body {
font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
margin: 0;
padding: 20px;
background-color: #f5f5f5;
}
.container {
max-width: 1200px;
margin: 0 auto;
}
.header {
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
color: white;
padding: 20px;
border-radius: 10px;
margin-bottom: 20px;
text-align: center;
}
.stats-grid {
display: grid;
grid-template-columns: repeat(auto-fit, minmax(250px, 1fr));
gap: 20px;
margin-bottom: 20px;
}
.stat-card {
background: white;
padding: 20px;
border-radius: 10px;
box-shadow: 0 2px 10px rgba(0,0,0,0.1);
text-align: center;
}
.stat-number {
font-size: 2.5em;
font-weight: bold;
color: #667eea;
margin: 10px 0;
}
.stat-label {
color: #666;
font-size: 0.9em;
}
.section {
background: white;
padding: 20px;
border-radius: 10px;
box-shadow: 0 2px 10px rgba(0,0,0,0.1);
margin-bottom: 20px;
}
.section h3 {
margin-top: 0;
color: #333;
border-bottom: 2px solid #667eea;
padding-bottom: 10px;
}
.server-status {
display: grid;
grid-template-columns: repeat(auto-fit, minmax(300px, 1fr));
gap: 15px;
margin-top: 15px;
}
.server-card {
border: 1px solid #ddd;
border-radius: 8px;
padding: 15px;
background: #fafafa;
}
.server-card.online {
border-color: #4CAF50;
background: #f1f8e9;
}
.server-card.offline {
border-color: #f44336;
background: #ffebee;
}
.status-indicator {
display: inline-block;
width: 12px;
height: 12px;
border-radius: 50%;
margin-right: 8px;
}
.status-online { background-color: #4CAF50; }
.status-offline { background-color: #f44336; }
.refresh-btn {
background: #667eea;
color: white;
border: none;
padding: 10px 20px;
border-radius: 5px;
cursor: pointer;
font-size: 14px;
margin-bottom: 20px;
}
.refresh-btn:hover {
background: #5a6fd8;
}
.task-list {
max-height: 400px;
overflow-y: auto;
}
.task-item {
border: 1px solid #ddd;
border-radius: 5px;
padding: 10px;
margin: 5px 0;
background: #f9f9f9;
}
.task-status {
display: inline-block;
padding: 2px 8px;
border-radius: 3px;
font-size: 0.8em;
color: white;
}
.status-pending { background: #ff9800; }
.status-running { background: #2196F3; }
.status-completed { background: #4CAF50; }
.status-failed { background: #f44336; }
</style>
</head>
<body>
<div class="container">
<div class="header">
<h1>🚀 ComfyUI 工作流服务监控</h1>
<p>实时监控系统状态、任务队列和服务器健康状态</p>
</div>
<button class="refresh-btn" onclick="refreshData()">🔄 刷新数据</button>
<div class="stats-grid">
<div class="stat-card">
<div class="stat-number" id="cpu-usage">--</div>
<div class="stat-label">CPU 使用率</div>
</div>
<div class="stat-card">
<div class="stat-number" id="memory-usage">--</div>
<div class="stat-label">内存使用率</div>
</div>
<div class="stat-card">
<div class="stat-number" id="running-tasks">--</div>
<div class="stat-label">运行中任务</div>
</div>
<div class="stat-card">
<div class="stat-number" id="pending-tasks">--</div>
<div class="stat-label">等待中任务</div>
</div>
</div>
<div class="section">
<h3>🖥️ 服务器状态</h3>
<div class="server-status" id="server-status">
<!-- 服务器状态将在这里动态加载 -->
</div>
</div>
<div class="section">
<h3>📋 最近任务</h3>
<div class="task-list" id="recent-tasks">
<!-- 最近任务将在这里动态加载 -->
</div>
</div>
</div>
<script>
async function refreshData() {
try {
// 获取系统状态
const systemStats = await fetch('/monitor/system-stats').then(r => r.json());
document.getElementById('cpu-usage').textContent = systemStats.cpu_percent + '%';
document.getElementById('memory-usage').textContent = systemStats.memory_percent + '%';
// 获取任务状态
const taskStats = await fetch('/monitor/task-stats').then(r => r.json());
document.getElementById('running-tasks').textContent = taskStats.running_tasks;
document.getElementById('pending-tasks').textContent = taskStats.pending_tasks;
// 获取服务器状态
const serverStatus = await fetch('/monitor/server-status').then(r => r.json());
updateServerStatus(serverStatus);
// 获取最近任务
const recentTasks = await fetch('/monitor/recent-tasks').then(r => r.json());
updateRecentTasks(recentTasks);
} catch (error) {
console.error('刷新数据失败:', error);
}
}
function updateServerStatus(servers) {
const container = document.getElementById('server-status');
container.innerHTML = '';
servers.forEach(server => {
const statusClass = server.status === 'online' ? 'online' : 'offline';
const statusColor = server.status === 'online' ? 'status-online' : 'status-offline';
container.innerHTML += `
<div class="server-card ${statusClass}">
<h4>${server.http_url}</h4>
<p><span class="status-indicator ${statusColor}"></span>${server.status}</p>
<p><strong>输入目录:</strong> ${server.input_dir}</p>
<p><strong>输出目录:</strong> ${server.output_dir}</p>
</div>
`;
});
}
function updateRecentTasks(tasks) {
const container = document.getElementById('recent-tasks');
container.innerHTML = '';
tasks.forEach(task => {
const statusClass = 'status-' + task.status.toLowerCase();
container.innerHTML += `
<div class="task-item">
<div style="display: flex; justify-content: space-between; align-items: center;">
<div>
<strong>${task.workflow_name}</strong>
<br>
<small>ID: ${task.workflow_run_id}</small>
</div>
<span class="task-status ${statusClass}">${task.status}</span>
</div>
<div style="margin-top: 5px; color: #666;">
<small>创建时间: ${new Date(task.created_at).toLocaleString()}</small>
</div>
</div>
`;
});
}
// 页面加载时自动刷新数据
document.addEventListener('DOMContentLoaded', refreshData);
// 每30秒自动刷新一次
setInterval(refreshData, 30000);
</script>
</body>
</html>
"""
return HTMLResponse(content=html_content)
@monitor_router.get("/system-stats")
async def get_system_stats() -> Dict[str, Any]:
"""获取系统统计信息"""
try:
cpu_percent = psutil.cpu_percent(interval=1)
memory = psutil.virtual_memory()
disk = psutil.disk_usage('/')
return {
"cpu_percent": round(cpu_percent, 1),
"memory_percent": round(memory.percent, 1),
"memory_used_gb": round(memory.used / (1024**3), 2),
"memory_total_gb": round(memory.total / (1024**3), 2),
"disk_percent": round(disk.percent, 1),
"disk_used_gb": round(disk.used / (1024**3), 2),
"disk_total_gb": round(disk.total / (1024**3), 2),
"timestamp": datetime.now().isoformat()
}
except Exception as e:
raise HTTPException(status_code=500, detail=f"获取系统统计信息失败: {str(e)}")
@monitor_router.get("/task-stats")
async def get_task_stats() -> Dict[str, Any]:
"""获取任务统计信息"""
try:
# 获取最近24小时的任务统计
end_time = datetime.now()
start_time = end_time - timedelta(hours=24)
recent_runs = await get_workflow_runs_recent(start_time, end_time)
# 统计各种状态的任务数量
status_counts = {}
for run in recent_runs:
status = run.get('status', 'unknown')
status_counts[status] = status_counts.get(status, 0) + 1
return {
"running_tasks": status_counts.get('running', 0),
"pending_tasks": status_counts.get('pending', 0),
"completed_tasks": status_counts.get('completed', 0),
"failed_tasks": status_counts.get('failed', 0),
"total_tasks_24h": len(recent_runs),
"timestamp": datetime.now().isoformat()
}
except Exception as e:
raise HTTPException(status_code=500, detail=f"获取任务统计信息失败: {str(e)}")
@monitor_router.get("/server-status")
async def get_server_status() -> List[Dict[str, Any]]:
"""获取服务器状态信息"""
try:
import aiohttp
import asyncio
async def check_server_status(server):
try:
timeout = aiohttp.ClientTimeout(total=5)
async with aiohttp.ClientSession(timeout=timeout) as session:
async with session.get(f"{server.http_url}/system_stats") as response:
if response.status == 200:
return {
"http_url": server.http_url,
"ws_url": server.ws_url,
"input_dir": server.input_dir,
"output_dir": server.output_dir,
"status": "online",
"response_time": response.headers.get('X-Response-Time', 'N/A')
}
else:
return {
"http_url": server.http_url,
"ws_url": server.ws_url,
"input_dir": server.input_dir,
"output_dir": server.output_dir,
"status": "offline",
"error": f"HTTP {response.status}"
}
except Exception as e:
return {
"http_url": server.http_url,
"ws_url": server.ws_url,
"input_dir": server.input_dir,
"output_dir": server.output_dir,
"status": "offline",
"error": str(e)
}
# 并发检查所有服务器状态
tasks = [check_server_status(server) for server in settings.SERVERS]
results = await asyncio.gather(*tasks, return_exceptions=True)
# 过滤掉异常结果
valid_results = []
for result in results:
if isinstance(result, dict):
valid_results.append(result)
return valid_results
except Exception as e:
raise HTTPException(status_code=500, detail=f"获取服务器状态失败: {str(e)}")
@monitor_router.get("/recent-tasks")
async def get_recent_tasks(limit: int = 10) -> List[Dict[str, Any]]:
"""获取最近的任务列表"""
try:
end_time = datetime.now()
start_time = end_time - timedelta(hours=24)
recent_runs = await get_workflow_runs_recent(start_time, end_time)
# 限制返回数量
limited_runs = recent_runs[:limit]
# 格式化返回数据
formatted_runs = []
for run in limited_runs:
formatted_runs.append({
"workflow_run_id": run.get('workflow_run_id'),
"workflow_name": run.get('workflow_name'),
"status": run.get('status', 'unknown'),
"created_at": run.get('created_at'),
"updated_at": run.get('updated_at'),
"api_spec": run.get('api_spec')
})
return formatted_runs
except Exception as e:
raise HTTPException(status_code=500, detail=f"获取最近任务失败: {str(e)}")
@monitor_router.get("/health")
async def health_check() -> Dict[str, Any]:
"""健康检查端点"""
try:
# 检查数据库连接
from workflow_service.database.connection import get_db
db = get_db()
# 检查配置
servers_count = len(settings.SERVERS)
return {
"status": "healthy",
"timestamp": datetime.now().isoformat(),
"version": "1.0.0",
"database": "connected",
"servers_configured": servers_count,
"uptime": "N/A" # 可以添加启动时间跟踪
}
except Exception as e:
return {
"status": "unhealthy",
"timestamp": datetime.now().isoformat(),
"error": str(e)
}