add gpuusage的逻辑
Change-Id: Ie1634a7d58ab94b05a3ebd7de87d362ca9be2286
diff --git a/Merge/back_rhj/app.py b/Merge/back_rhj/app.py
index df7a598..967b921 100644
--- a/Merge/back_rhj/app.py
+++ b/Merge/back_rhj/app.py
@@ -1,6 +1,47 @@
from app import create_app
+from flask_cors import CORS
+import os
+import psutil
+import time
+import GPUtil
+from flask import g, request
+from sqlalchemy import create_engine
+from sqlalchemy.orm import sessionmaker
+from config import Config
+from app.utils.gpuwriter import GpuWriter
app = create_app()
+CORS(app,
+ resources={r"/*": {"origins": "*"}},
+ supports_credentials=True,
+ allow_headers=["Content-Type", "Authorization"]
+)
+
+proc = psutil.Process(os.getpid())
+
+@app.before_request
+def before_request():
+ g.start_time = time.time()
+ g.start_cpu = proc.cpu_times()
+ g.start_mem = proc.memory_info()
+
+@app.after_request
+def after_request(response):
+ try:
+ # 记录GPU使用情况
+ gpu_writer = GpuWriter(Config.SQLURL)
+ gpus = GPUtil.getGPUs()
+
+ for gpu in gpus:
+ gpu_writer.write_gpu_usage(
+ gpu_id=gpu.id,
+ gpu_usage=gpu.load * 100, # 转换为百分比
+ gpu_memory_usage=gpu.memoryUsed # MB
+ )
+ except Exception as e:
+ print(f"GPU使用情况记录失败: {e}")
+
+ return response
if __name__ == "__main__":
app.run(debug=True,port=8082,host='0.0.0.0')
\ No newline at end of file
diff --git a/Merge/back_rhj/app/__pycache__/__init__.cpython-310.pyc b/Merge/back_rhj/app/__pycache__/__init__.cpython-310.pyc
index bc775fd..08f5017 100644
--- a/Merge/back_rhj/app/__pycache__/__init__.cpython-310.pyc
+++ b/Merge/back_rhj/app/__pycache__/__init__.cpython-310.pyc
Binary files differ
diff --git a/Merge/back_rhj/app/__pycache__/routes.cpython-310.pyc b/Merge/back_rhj/app/__pycache__/routes.cpython-310.pyc
index f5aa9b3..46bb4b4 100644
--- a/Merge/back_rhj/app/__pycache__/routes.cpython-310.pyc
+++ b/Merge/back_rhj/app/__pycache__/routes.cpython-310.pyc
Binary files differ
diff --git a/Merge/back_rhj/app/models/__pycache__/gpu_usage.cpython-310.pyc b/Merge/back_rhj/app/models/__pycache__/gpu_usage.cpython-310.pyc
new file mode 100644
index 0000000..79f092a
--- /dev/null
+++ b/Merge/back_rhj/app/models/__pycache__/gpu_usage.cpython-310.pyc
Binary files differ
diff --git a/Merge/back_rhj/app/models/gpu_usage.py b/Merge/back_rhj/app/models/gpu_usage.py
new file mode 100644
index 0000000..0ac5cfa
--- /dev/null
+++ b/Merge/back_rhj/app/models/gpu_usage.py
@@ -0,0 +1,18 @@
+from sqlalchemy import Column, Integer, DECIMAL, TIMESTAMP, text
+from sqlalchemy.ext.declarative import declarative_base
+from sqlalchemy.sql import func
+
+Base = declarative_base()
+
+class GpuUsage(Base):
+ __tablename__ = 'gpu_usage'
+ __table_args__ = {
+ 'mysql_engine': 'InnoDB',
+ 'mysql_charset': 'utf8mb4',
+ 'comment': 'GPU 使用情况表'
+ }
+
+ gpu_id = Column(Integer, primary_key=True, nullable=False, comment='GPU 编号')
+ gpu_usage = Column(DECIMAL(5, 2), nullable=False, comment='GPU 使用率，单位：百分比')
+ gpu_memory_usage = Column(Integer, nullable=False, comment='GPU 内存用量，单位：MB')
+ created_at = Column(TIMESTAMP, nullable=False, server_default=text('CURRENT_TIMESTAMP'), comment='记录时间戳')
diff --git a/Merge/back_rhj/app/models/recall/__pycache__/swing_recall.cpython-310.pyc b/Merge/back_rhj/app/models/recall/__pycache__/swing_recall.cpython-310.pyc
index 5cd2aee..f50ef90 100644
--- a/Merge/back_rhj/app/models/recall/__pycache__/swing_recall.cpython-310.pyc
+++ b/Merge/back_rhj/app/models/recall/__pycache__/swing_recall.cpython-310.pyc
Binary files differ
diff --git a/Merge/back_rhj/app/routes.py b/Merge/back_rhj/app/routes.py
index 6f159ca..0b60624 100644
--- a/Merge/back_rhj/app/routes.py
+++ b/Merge/back_rhj/app/routes.py
@@ -2,6 +2,7 @@
from .functions.FAuth import FAuth
from app.services.recommendation_service import RecommendationService
from app.utils.graph_build import build_user_post_graph
+from app.utils.gpuwriter import GpuWriter
from sqlalchemy import create_engine
from sqlalchemy.orm import sessionmaker
from config import Config
@@ -409,3 +410,37 @@
'success': False,
'message': f'推荐获取失败: {str(e)}'
}), 500
+
+@main.route('/gpu-usage', methods=['GET'])
+@token_required
+def get_gpu_usage(current_user):
+ """获取最新GPU使用情况接口"""
+ try:
+ limit = request.args.get('limit', 100, type=int)
+
+ # 限制最大获取条数
+ if limit > 1000:
+ limit = 1000
+
+ print(f"获取最新 {limit} 条GPU使用记录")
+
+ # 创建GpuWriter实例并获取数据
+ gpu_writer = GpuWriter(Config.SQLURL)
+ gpu_records = gpu_writer.get_latest_gpu_usage(limit)
+
+ return jsonify({
+ 'success': True,
+ 'data': {
+ 'records': gpu_records,
+ 'count': len(gpu_records),
+ 'limit': limit
+ },
+ 'message': 'GPU使用情况获取成功'
+ }), 200
+
+ except Exception as e:
+ print(f"获取GPU使用情况错误: {str(e)}")
+ return jsonify({
+ 'success': False,
+ 'message': f'获取GPU使用情况失败: {str(e)}'
+ }), 500
diff --git a/Merge/back_rhj/app/services/__pycache__/lightgcn_scorer.cpython-310.pyc b/Merge/back_rhj/app/services/__pycache__/lightgcn_scorer.cpython-310.pyc
index 495528d..c98aaf5 100644
--- a/Merge/back_rhj/app/services/__pycache__/lightgcn_scorer.cpython-310.pyc
+++ b/Merge/back_rhj/app/services/__pycache__/lightgcn_scorer.cpython-310.pyc
Binary files differ
diff --git a/Merge/back_rhj/app/services/__pycache__/recommendation_service.cpython-310.pyc b/Merge/back_rhj/app/services/__pycache__/recommendation_service.cpython-310.pyc
index 88ab960..1424b42 100644
--- a/Merge/back_rhj/app/services/__pycache__/recommendation_service.cpython-310.pyc
+++ b/Merge/back_rhj/app/services/__pycache__/recommendation_service.cpython-310.pyc
Binary files differ
diff --git a/Merge/back_rhj/app/utils/__pycache__/gpuwriter.cpython-310.pyc b/Merge/back_rhj/app/utils/__pycache__/gpuwriter.cpython-310.pyc
new file mode 100644
index 0000000..f86da67
--- /dev/null
+++ b/Merge/back_rhj/app/utils/__pycache__/gpuwriter.cpython-310.pyc
Binary files differ
diff --git a/Merge/back_rhj/app/utils/gpuwriter.py b/Merge/back_rhj/app/utils/gpuwriter.py
new file mode 100644
index 0000000..1e521f3
--- /dev/null
+++ b/Merge/back_rhj/app/utils/gpuwriter.py
@@ -0,0 +1,68 @@
+from sqlalchemy.orm import sessionmaker
+from sqlalchemy import create_engine
+from app.models.gpu_usage import GpuUsage
+import logging
+
+class GpuWriter:
+ def __init__(self, database_url):
+ """
+ 初始化 GpuWriter
+ :param database_url: 数据库连接URL
+ """
+ self.engine = create_engine(database_url)
+ self.Session = sessionmaker(bind=self.engine)
+
+ def write_gpu_usage(self, gpu_id, gpu_usage, gpu_memory_usage):
+ """
+ 写入GPU使用情况到数据库
+ :param gpu_id: GPU编号
+ :param gpu_usage: GPU使用率（百分比）
+ :param gpu_memory_usage: GPU内存用量（MB）
+ :return: bool 是否写入成功
+ """
+ session = self.Session()
+ try:
+ gpu_record = GpuUsage(
+ gpu_id=gpu_id,
+ gpu_usage=gpu_usage,
+ gpu_memory_usage=gpu_memory_usage
+ )
+ session.add(gpu_record)
+ session.commit()
+ return True
+ except Exception as e:
+ session.rollback()
+ logging.error(f"写入GPU使用情况失败: {e}")
+ return False
+ finally:
+ session.close()
+
+ def get_latest_gpu_usage(self, limit=100):
+ """
+ 获取最新的GPU使用情况记录
+ :param limit: 获取记录条数，默认100条
+ :return: list 包含GPU使用记录的字典列表
+ """
+ session = self.Session()
+ try:
+ records = session.query(GpuUsage)\
+ .order_by(GpuUsage.timestamp.desc())\
+ .limit(limit)\
+ .all()
+
+ result = []
+ for record in records:
+ result.append({
+ 'id': record.id,
+ 'gpu_id': record.gpu_id,
+ 'gpu_usage': record.gpu_usage,
+ 'gpu_memory_usage': record.gpu_memory_usage,
+ 'timestamp': record.timestamp.isoformat() if record.timestamp else None
+ })
+
+ return result
+ except Exception as e:
+ logging.error(f"获取GPU使用情况失败: {e}")
+ return []
+ finally:
+ session.close()
diff --git a/Merge/front/src/api/posts_trm.js b/Merge/front/src/api/posts_trm.js
index d6303a7..8f7e877 100644
--- a/Merge/front/src/api/posts_trm.js
+++ b/Merge/front/src/api/posts_trm.js
@@ -209,4 +209,30 @@
}
console.log('Normalized sys cost list:', list)
return list
+}
+
+/**
+ * 获取GPU使用情况数据
+ * GET /gpu-usage
+ * @param {number} limit 获取记录条数，默认100条
+ * @returns Promise<[ {id, gpu_id, gpu_usage, gpu_memory_usage, timestamp}, … ]>
+ */
+export async function fetchGpuUsage(limit = 100) {
+ const token = localStorage.getItem('token')
+ const res = await fetch(`http://10.126.59.25:8082/gpu-usage?limit=${limit}`, {
+ method: 'GET',
+ headers: {
+ 'Content-Type': 'application/json',
+ 'Authorization': `Bearer ${token}`
+ }
+ })
+ if (!res.ok) throw new Error(`fetchGpuUsage: ${res.status}`)
+ const json = await res.json()
+ console.log('fetchGpuUsage response:', json)
+
+ if (json.success === false) {
+ throw new Error(json.message || 'Failed to fetch GPU usage')
+ }
+
+ return json.data?.records || []
}
\ No newline at end of file
diff --git a/Merge/front/src/components/PerformanceLogs.js b/Merge/front/src/components/PerformanceLogs.js
index f87ef7a..67518de 100644
--- a/Merge/front/src/components/PerformanceLogs.js
+++ b/Merge/front/src/components/PerformanceLogs.js
@@ -3,16 +3,19 @@
ResponsiveContainer, LineChart, Line,
XAxis, YAxis, Tooltip, CartesianGrid, Legend
} from 'recharts';
-import { fetchSysCost } from '../api/posts_trm';
+import { fetchSysCost, fetchGpuUsage } from '../api/posts_trm';
function PerformanceLogs({ userId }) {
const [data, setData] = useState([]);
+ const [gpuData, setGpuData] = useState([]);
const [loading, setLoading] = useState(true);
const [unauthorized, setUnauthorized] = useState(false);
useEffect(() => {
setLoading(true);
- fetchSysCost(userId)
+
+ // 获取系统性能数据
+ const fetchSystemData = fetchSysCost(userId)
.then(result => {
// 检查是否是权限错误
if (result && result.status === 'error' && result.message === 'Unauthorized') {
@@ -49,10 +52,59 @@
setUnauthorized(true);
setData([]);
}
+ });
+
+ // 获取GPU数据
+ const fetchGpuData = fetchGpuUsage(100)
+ .then(result => {
+ console.log('GPU data:', result);
+ const processedData = processGpuData(result);
+ setGpuData(processedData);
})
+ .catch(err => {
+ console.error('fetchGpuUsage error:', err);
+ });
+
+ Promise.all([fetchSystemData, fetchGpuData])
.finally(() => setLoading(false));
}, [userId]);
+ // 处理GPU数据，按时间戳分组
+ const processGpuData = (rawData) => {
+ const timeMap = new Map();
+
+ rawData.forEach(record => {
+ const timestamp = new Date(record.timestamp).getTime();
+ if (!timeMap.has(timestamp)) {
+ timeMap.set(timestamp, { timestamp });
+ }
+ const timeEntry = timeMap.get(timestamp);
+ timeEntry[`gpu${record.gpu_id}_usage`] = record.gpu_usage;
+ timeEntry[`gpu${record.gpu_id}_memory`] = record.gpu_memory_usage;
+ });
+
+ return Array.from(timeMap.values()).sort((a, b) => a.timestamp - b.timestamp);
+ };
+
+ // 获取GPU颜色
+ const getGpuColors = () => {
+ return ['#8884d8', '#82ca9d', '#ffc658', '#ff7300', '#8dd1e1', '#d084d0', '#ffb347', '#87ceeb'];
+ };
+
+ // 获取唯一的GPU ID列表
+ const getGpuIds = () => {
+ const ids = new Set();
+ gpuData.forEach(entry => {
+ Object.keys(entry).forEach(key => {
+ if (key.includes('gpu') && key.includes('_usage')) {
+ const gpuId = key.replace('_usage', '').replace('gpu', '');
+ ids.add(gpuId);
+ }
+ });
+ });
+ return Array.from(ids).sort((a, b) => parseInt(a) - parseInt(b));
+ };
+
if (loading) {
return (
<section className="dashboard-performance">
@@ -114,6 +166,70 @@
return (
<section className="dashboard-performance">
+ {/* GPU使用率图表 */}
+ <div style={{ marginBottom: '30px' }}>
+ <h3>GPU 使用率</h3>
+ <ResponsiveContainer width="100%" height={300}>
+ <LineChart data={gpuData} margin={{ top: 20, right: 30, left: 20, bottom: 5 }}>
+ <CartesianGrid strokeDasharray="3 3" />
+ <XAxis
+ dataKey="timestamp"
+ type="number"
+ domain={['dataMin', 'dataMax']}
+ tickFormatter={ts => new Date(ts).toLocaleTimeString()}
+ />
+ <YAxis domain={[0, 100]} label={{ value: '使用率 (%)', angle: -90, position: 'insideLeft' }} />
+ <Tooltip
+ formatter={(val, name) => [`${val?.toFixed(2)}%`, name]}
+ labelFormatter={ts => new Date(ts).toLocaleString()}
+ />
+ <Legend />
+ {getGpuIds().map((gpuId, index) => (
+ <Line
+ key={`gpu${gpuId}_usage`}
+ type="monotone"
+ dataKey={`gpu${gpuId}_usage`}
+ stroke={getGpuColors()[index % getGpuColors().length]}
+ name={`GPU ${gpuId} 使用率`}
+ connectNulls={false}
+ />
+ ))}
+ </LineChart>
+ </ResponsiveContainer>
+ </div>
+
+ {/* GPU内存占用图表 */}
+ <div style={{ marginBottom: '30px' }}>
+ <h3>GPU 内存占用</h3>
+ <ResponsiveContainer width="100%" height={300}>
+ <LineChart data={gpuData} margin={{ top: 20, right: 30, left: 20, bottom: 5 }}>
+ <CartesianGrid strokeDasharray="3 3" />
+ <XAxis
+ dataKey="timestamp"
+ type="number"
+ domain={['dataMin', 'dataMax']}
+ tickFormatter={ts => new Date(ts).toLocaleTimeString()}
+ />
+ <YAxis domain={[0, 'auto']} label={{ value: '内存 (MB)', angle: -90, position: 'insideLeft' }} />
+ <Tooltip
+ formatter={(val, name) => [`${val?.toFixed(2)} MB`, name]}
+ labelFormatter={ts => new Date(ts).toLocaleString()}
+ />
+ <Legend />
+ {getGpuIds().map((gpuId, index) => (
+ <Line
+ key={`gpu${gpuId}_memory`}
+ type="monotone"
+ dataKey={`gpu${gpuId}_memory`}
+ stroke={getGpuColors()[index % getGpuColors().length]}
+ name={`GPU ${gpuId} 内存`}
+ connectNulls={false}
+ />
+ ))}
+ </LineChart>
+ </ResponsiveContainer>
+ </div>
+
{/* 响应时间图表 */}
<div style={{ marginBottom: '30px' }}>
<h3>响应时间</h3>