Skip to content
获取系统资源使用率
方案一
软件的安装卸载：
使用文档：
GPU监控
nano /etc/systemd/system/beszel-agent.service
[Unit]
Description=Beszel Agent Service
Wants=network-online.target
After=network-online.target
[Service]
DeviceAllow=/dev/nvidiactl rw
DeviceAllow=/dev/nvidia0 rw
Environment="GPU=true"
Environment="PORT=45876"
Environment="KEY=ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIAspPzu1E73R9gqdOFcLsVdycKfzww0GcLHoma3dxmQo"
# Environment="EXTRA_FILESYSTEMS=sdb"
ExecStart=/opt/beszel-agent/beszel-agent
User=beszel
Restart=on-failure
RestartSec=5
StateDirectory=beszel-agent
# Security/sandboxing settings
KeyringMode=private
LockPersonality=yes
NoNewPrivileges=yes
PrivateTmp=yes
ProtectClock=yes
ProtectHome=read-only
ProtectHostname=yes
ProtectKernelLogs=yes
ProtectKernelTunables=yes
ProtectSystem=strict
RemoveIPC=yes
RestrictSUIDSGID=true
SystemCallArchitectures=native
[Install]
WantedBy=multi-user.target
systemctl daemon-reload
sudo systemctl restart beszel-agent
解决GPU不显示
卸载：
./install-agent.sh -u
api 使用文档：
import PocketBase from 'pocketbase'
const pb = new PocketBase('http://10.100.1.125:8090')
// 作为普通用户进行身份验证
const userData = await pb.collection('users').authWithPassword('sspu@sspu.com', 'sspu106106')
// 列出并过滤系统记录
const systems = await pb.collection('systems').getList(1, 20, {
filter: 'status = "up" && created > "2025-02-01 10:00:00"',
})
console.log(systems)
获取详细情况
import PocketBase from 'pocketbase';
const pb = new PocketBase('http://10.100.1.125:8090');
async function getSystemStats(recordId) {
const record = await pb.collection('system_stats').getOne(recordId, {
expand: 'cpu_usage,memory_usage,disk_usage,network_usage',
});
console.log(record);
}
getSystemStats('YOUR_RECORD_ID');
胡乱回答
方案二python
预先安装
a：
pip install flask-cors
192.168.10.210 g1 node10
192.168.10.211 g2 node11
192.168.10.212 g3 node12
c
from flask import Flask, request, jsonify
from flask_cors import CORS
import threading
import time
import requests
import psutil
import nvidia_smi
app = Flask(__name__)
CORS(app)
# 存储所有机器的状态数据 (包括 A, B, C)
machine_status = {}
# 内网机器 A、B 的 IP 地址（请替换为实际 IP）
INTERNAL_MACHINES = {
"A": "http://192.168.1.101:5000",
"B": "http://192.168.1.102:5000"
}
def format_size(bytes, unit='GB'):
"""格式化存储单位"""
units = {'KB': 1e3, 'MB': 1e6, 'GB': 1e9, 'TB': 1e12}
return round(bytes / units[unit], 2)
def get_gpu_status():
"""获取 GPU 资源使用情况"""
try:
nvidia_smi.nvmlInit()
except Exception as e:
print("nvidia_smi 初始化失败：", e)
return []
gpus = []
try:
device_count = nvidia_smi.nvmlDeviceGetCount()
for i in range(device_count):
try:
handle = nvidia_smi.nvmlDeviceGetHandleByIndex(i)
mem_info = nvidia_smi.nvmlDeviceGetMemoryInfo(handle)
util_rate = nvidia_smi.nvmlDeviceGetUtilizationRates(handle)
temp = nvidia_smi.nvmlDeviceGetTemperature(handle, nvidia_smi.NVML_TEMPERATURE_GPU)
name = nvidia_smi.nvmlDeviceGetName(handle).decode('utf-8')
gpus.append({
'id': i,
'name': name,
'memory_free': format_size(mem_info.free),
'memory_total': format_size(mem_info.total),
'usage_percent': round(mem_info.used / mem_info.total * 100, 2),
'utilization_gpu': util_rate.gpu,
'utilization_memory': util_rate.memory,
'temperature': temp
})
except Exception as e:
print(f"获取 GPU {i} 状态失败：", e)
except Exception as e:
print("获取 GPU 数量失败：", e)
nvidia_smi.nvmlShutdown()
return gpus
def get_system_status():
"""获取系统资源使用情况"""
cpu_usage = psutil.cpu_percent(interval=1)
memory = psutil.virtual_memory()
disk = psutil.disk_usage('/')
net = psutil.net_io_counters()
return {
'cpu_usage': cpu_usage,
'cpu_cores': psutil.cpu_count(logical=False),
'cpu_threads': psutil.cpu_count(logical=True),
'memory_total': format_size(memory.total),
'memory_used': format_size(memory.used),
'memory_free': format_size(memory.free),
'memory_percent': memory.percent,
'disk_total': format_size(disk.total, 'TB'),
'disk_used': format_size(disk.used, 'GB'),
'disk_free': format_size(disk.free, 'GB'),
'disk_percent': disk.percent,
'net_sent': format_size(net.bytes_sent, 'MB'),
'net_recv': format_size(net.bytes_recv, 'MB')
}
def collect_local_status():
"""
定时采集公网机器 C 自身的状态，并存储到 machine_status。
"""
while True:
try:
data = {
'machine': 'C',
'system': get_system_status(),
'gpu': get_gpu_status()
}
machine_status['C'] = data
except Exception as e:
print("采集 C 机器状态失败：", e)
# 每隔 10 秒更新一次数据
time.sleep(10)
def fetch_internal_status():
"""
定时从 A、B 机器获取状态数据，并存储到 machine_status。
"""
while True:
for name, url in INTERNAL_MACHINES.items():
try:
response = requests.get(f"{url}/full_status", timeout=5)
if response.status_code == 200:
machine_status[name] = response.json()
print(f"成功获取 {name} 的状态数据")
else:
print(f"获取 {name} 的状态失败，状态码：{response.status_code}")
except requests.RequestException as e:
print(f"请求 {name} 失败：", e)
# 每隔 10 秒获取一次数据
time.sleep(10)
@app.route('/update_status', methods=['POST'])
def update_status():
"""
允许 A 和 B 主动推送状态数据，C 机器会存储这些数据。
"""
data = request.get_json()
machine_name = data.get('machine')
if machine_name:
machine_status[machine_name] = data
return jsonify({"message": "Status updated successfully"}), 200
return jsonify({"error": "Invalid data"}), 400
@app.route('/full_status', methods=['GET'])
def full_status():
"""
返回所有机器的状态数据，包括 C 本身和 A、B 发送的数据。
"""
return jsonify(machine_status)
if __name__ == '__main__':
# 启动采集 C 机器自身状态的线程
threading.Thread(target=collect_local_status, daemon=True).start()
# 启动获取 A、B 状态数据的线程
threading.Thread(target=fetch_internal_status, daemon=True).start()
# 运行 Web 服务器
app.run(host='0.0.0.0', port=5000)
a-b
import requests
import time
C_SERVER_URL = "http://公网IP:5000/update_status"
def send_status():
while True:
data = {
'machine': 'A', # A 机器的名字
'system': get_system_status(),
'gpu': get_gpu_status()
}
try:
requests.post(C_SERVER_URL, json=data, timeout=5)
except requests.RequestException as e:
print("上传状态失败：", e)
time.sleep(10) # 每 10 秒上传一次
send_status()
Want to print your doc?
This is not the way.
This is not the way.
Try clicking the ⋯ next to your doc name or using a keyboard shortcut (
CtrlP
) instead.