import subprocess import os import threading import time import yaml from datetime import datetime import signal import psutil import glob import re import pytz import requests import json DATA_JSON = os.environ.get('DATA_JSON', '') def assign_vars_from_json(config: dict): for key, value in config.items(): globals()[key] = value if DATA_JSON: # 如果 DATA_JSON 有值 try: config_dict = json.loads(DATA_JSON) # 解析 JSON 字符串 assign_vars_from_json(config_dict) except json.JSONDecodeError: raise ValueError("DATA_JSON 环境变量不是合法的 JSON 字符串") else: # 如果 DATA_JSON 为空，就从单独的环境变量里取 N_PORT = os.environ.get('N_PORT', '8008') # N Z端口 默认8008 ARGO_PORT = os.environ.get('ARGO_PORT', '8009') # Argo（nginx的反代端口）固定隧道端口,留空默认8009 ARGO_DOMAIN = os.environ.get('ARGO_DOMAIN', '') # Argo固定隧道域名,留空即使用临时隧道 ARGO_AUTH = os.environ.get('ARGO_AUTH', '') # Argo固定隧道密钥,留空即使用临时隧道 DASHBOARD_VERSION = os.environ.get('DASHBOARD_VERSION', 'v1.13.2')# 指定面板的版本，以 v0.00.00 的格式，后续将固定在该版本不会升级，不填则使用默认的 v1.13.2 NZV1_VERSION = os.environ.get('NZV1_VERSION', 'v1.13.1')# 哪吒V1的版本默认v1.13.1 BACKUP_TIME = os.environ.get('BACKUP_TIME', '3600')# 备份时间 10800秒 2小时 RESTART_TIME = os.environ.get('RESTART_TIME', '14400')# 重启时间 4小时 14400 6小时 21600 HF_USER1 = os.environ.get('HF_USER1', '')# HF 备份仓库的用户名 HF_REPO = os.environ.get('HF_REPO', '')#HF 备份的Models仓库名 HF_EMAIL = os.environ.get('HF_EMAIL', '') #HF的邮箱 HF_TOKEN1 = os.environ.get('HF_TOKEN1', '')#HF备份账号的TOKEN HF_USER2 = os.environ.get('HF_USER2', '')# huggingface 用户名 HF_ID = os.environ.get('HF_ID', '')# huggingface space 名 HF_TOKEN2 = os.environ.get('HF_TOKEN2', '')# huggingface TOKEN #JUPYTER_TOKEN agent_config = { 'client_secret': 'MLcD6YnifhoY08B9n129UP5cg2139NYa', 'debug': False, 'disable_auto_update': True, 'disable_command_execute': False, 'disable_force_update': False, 'disable_nat': False, 'disable_send_query': False, 'gpu': False, 'insecure_tls': False, 'ip_report_period': 1800, 'report_delay': 3, 'self_update_period': 0, 'server': f'{ARGO_DOMAIN}:443', 'skip_connection_count': False, 'skip_procs_count': False, 'temperature': True, 'tls': True, 'use_gitee_to_upgrade': False, 'use_ipv6_country_code': False, 'uuid': '18a49016-bc2d-4be9-0ddb-5357fdbf0b3d' } dashboard_config = { 'admin_template': 'admin-dist', 'agent_secret_key': '', 'avg_ping_count': 2, 'cover': 1, 'https': {}, # 空字典 'install_host': f'{ARGO_DOMAIN}:443', 'ip_change_notification_group_id': 0, 'jwt_secret_key': '', 'jwt_timeout': 300, 'language': 'zh_CN', 'listen_port': f'{N_PORT}', 'location': 'Asia/Shanghai', 'site_name': '鸡子探针平台-柒蓝', 'tls': True, 'user_template': 'user-dist' } mime_types_content = """types { text/html html htm shtml; text/css css; text/javascript js; image/gif gif; image/jpeg jpeg jpg; image/png png; text/plain txt; application/json json; application/xml xml; application/octet-stream bin; }""" nginx_conf = """ # 全局配置 worker_processes auto; pid /tmp/nginx.pid; error_log /tmp/nginx_error.log; events { worker_connections 768; } http { sendfile on; tcp_nopush on; tcp_nodelay on; keepalive_timeout 65; types_hash_max_size 2048; # 修改 mime.types 路径 include /data/nginx1.24/mime.types; default_type application/octet-stream; # SSL配置 ssl_protocols TLSv1.2 TLSv1.3; ssl_prefer_server_ciphers on; access_log /tmp/nginx_access.log; gzip on; # 上游服务器配置 upstream dashboard { server 127.0.0.1:%s; keepalive 512; } # 服务器块 server { listen %s; listen [::]:%s; server_name %s; # 删除所有 real_ip 相关配置 underscores_in_headers on; # gRPC相关 location ^~ /proto.NezhaService/ { grpc_set_header Host $host; grpc_read_timeout 600s; grpc_send_timeout 600s; grpc_socket_keepalive on; client_max_body_size 10m; grpc_buffer_size 4m; grpc_pass grpc://dashboard; } # WebSocket相关 location ~* ^/api/v1/ws/(server|terminal|file)(.*)$ { proxy_set_header Host $host; proxy_set_header Origin https://$host; proxy_set_header Upgrade $http_upgrade; proxy_set_header Connection "upgrade"; proxy_read_timeout 3600s; proxy_send_timeout 3600s; proxy_pass http://127.0.0.1:%s; } # Web请求处理 location / { proxy_set_header Host $host; proxy_read_timeout 3600s; proxy_send_timeout 3600s; proxy_buffer_size 128k; proxy_buffers 4 256k; proxy_busy_buffers_size 256k; proxy_max_temp_file_size 0; proxy_pass http://127.0.0.1:%s; } # 安全头部 add_header X-Frame-Options SAMEORIGIN; add_header X-Content-Type-Options nosniff; add_header Strict-Transport-Security "max-age=31536000; includeSubDomains" always; } } """ % (N_PORT, ARGO_PORT, ARGO_PORT, ARGO_DOMAIN, N_PORT, N_PORT) stop_event = threading.Event() def kill_processes(): # 要结束的进程名列表 target_processes = [ 'cf', 'nv1', 'dv1', 'nginx' ] # 存储已结束的进程 killed_processes = [] # 遍历所有正在运行的进程 for proc in psutil.process_iter(['name']): try: # 检查进程名是否在目标列表中 if proc.info['name'] in target_processes: # 获取进程ID pid = proc.pid # 先尝试优雅地结束进程 proc.terminate() # 等待进程结束 try: proc.wait(timeout=3) except psutil.TimeoutExpired: # 如果进程未响应，强制杀死 proc.kill() killed_processes.append(f"{proc.info['name']} (PID: {pid})") except (psutil.NoSuchProcess, psutil.AccessDenied, psutil.ZombieProcess): pass # 打印已结束的进程 if killed_processes: print("已结束以下进程:") for process in killed_processes: print(process) else: print("未找到匹配的进程") kill_processes() def get_latest_local_package(directory, pattern='*.tar.gz'): try: # 构建完整的搜索路径 search_pattern = os.path.join(directory, pattern) # 查找所有匹配的文件 files = glob.glob(search_pattern) if not files: print("未找到匹配的 nezha-hf 压缩包") return None # 获取最新的文件 latest_file = max(files, key=os.path.getmtime) print(f"找到最新的包: {latest_file}") return latest_file except Exception as e: print(f"获取最新包时发生错误: {e}") return None def delete_huggingface_lfs_file(filename, repo_id, token): """ 通过Hugging Face API删除LFS文件记录 """ try: # 1. 查询LFS文件列表 url = f"https://huggingface.co/api/models/{repo_id}/lfs-files" headers = { "content-type": "application/json", "Authorization": f"Bearer {token}", "user-agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36" } response = requests.get(url, headers=headers) if response.status_code != 200: print(f"查询LFS文件失败: {response.status_code} - {response.text}") return False lfs_files = response.json() # 2. 查找匹配的文件 file_to_delete = None for file_info in lfs_files: if file_info.get('filename') == filename: file_to_delete = file_info break if not file_to_delete: print(f"未找到对应的LFS文件记录: {filename}") return False # 3. 删除LFS文件记录 file_oid = file_to_delete['fileOid'] delete_url = f"https://huggingface.co/api/models/{repo_id}/lfs-files/{file_oid}?rewriteHistory=true" delete_response = requests.delete(delete_url, headers=headers) if delete_response.status_code == 200: print(f"成功删除LFS文件记录: {filename} (OID: {file_oid})") return True else: print(f"删除LFS文件记录失败: {delete_response.status_code} - {delete_response.text}") return False except Exception as e: print(f"删除LFS文件记录时出错: {e}") return False def safe_git_cleanup(repo_path, files_to_remove): """ 安全的Git清理，不会影响现有的备份文件 """ try: original_dir = os.getcwd() os.chdir(repo_path) print(f"执行安全Git清理: {files_to_remove}") # 1. 首先确保工作目录是干净的 result = subprocess.run(['git', 'status', '--porcelain'], capture_output=True, text=True) if result.stdout.strip(): print("工作目录有未提交的更改，先提交...") subprocess.run(['git', 'add', '.'], capture_output=True) subprocess.run(['git', 'commit', '-m', '自动提交: 清理前的更改'], capture_output=True) # 2. 只从Git索引中删除指定的文件，但不影响工作目录中的文件 for filename in files_to_remove: if os.path.exists(filename): print(f"从Git索引中删除 {filename} (文件仍保留在工作目录)") subprocess.run(['git', 'rm', '--cached', filename], capture_output=True) else: print(f"文件 {filename} 不存在于工作目录，只清理Git引用") # 3. 提交删除操作 if files_to_remove: subprocess.run(['git', 'commit', '-m', f'清理已删除的文件: {", ".join(files_to_remove)}'], capture_output=True) # 4. 执行GC清理但不删除最近的文件 subprocess.run(['git', 'gc', '--auto'], capture_output=True) subprocess.run(['git', 'lfs', 'prune'], capture_output=True) print("安全Git清理完成") os.chdir(original_dir) return True except Exception as e: print(f"安全Git清理时出错: {e}") os.chdir(original_dir) return False def get_remote_lfs_files(repo_id, token): """ 获取远程所有的LFS文件列表 """ try: url = f"https://huggingface.co/api/models/{repo_id}/lfs-files" headers = { "content-type": "application/json", "Authorization": f"Bearer {token}", "user-agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36" } response = requests.get(url, headers=headers) if response.status_code == 200: return response.json() else: print(f"获取远程LFS文件失败: {response.status_code}") return [] except Exception as e: print(f"获取远程LFS文件时出错: {e}") return [] def cleanup_orphaned_lfs_references(repo_path, repo_id, token, keep_count=3): """ 清理孤儿LFS引用：删除远程存在但本地不存在的文件引用 """ try: original_dir = os.getcwd() os.chdir(repo_path) print("检查孤儿LFS引用...") # 获取远程LFS文件 remote_files = get_remote_lfs_files(repo_id, token) if not remote_files: print("无法获取远程LFS文件列表") return # 获取本地文件 local_files = set(glob.glob('*.tar.gz')) # 找出远程存在但本地不存在的文件 orphaned_files = [] for remote_file in remote_files: filename = remote_file.get('filename') if filename and filename not in local_files: orphaned_files.append(filename) if orphaned_files: print(f"发现孤儿LFS引用: {orphaned_files}") # 删除这些孤儿引用 for filename in orphaned_files: print(f"删除孤儿LFS引用: {filename}") delete_huggingface_lfs_file(filename, repo_id, token) print("孤儿LFS引用清理完成") os.chdir(original_dir) except Exception as e: print(f"清理孤儿LFS引用时出错: {e}") os.chdir(original_dir) def compress_folder(folder_path, output_dir, keep_count=3): try: # 确保输出目录存在 os.makedirs(output_dir, exist_ok=True) # 设置中国时区 china_tz = pytz.timezone('Asia/Shanghai') # 获取当前中国时间的时间戳 timestamp = str(int(datetime.now(china_tz).timestamp() * 1000)) output_path = os.path.join(output_dir, f'{timestamp}.tar.gz') # 获取所有压缩包 existing_archives = glob.glob(os.path.join(output_dir, '*.tar.gz')) # 安全地提取时间戳 def extract_timestamp(filename): match = re.search(r'(\d+)\.tar\.gz$', filename) return int(match.group(1)) if match else 0 files_to_cleanup = [] # 如果压缩包数量超过保留数量，删除最旧的 if len(existing_archives) >= keep_count: # 按时间戳排序（从小到大，最旧的在前面） existing_archives.sort(key=extract_timestamp) # 计算需要删除的数量 delete_count = len(existing_archives) - keep_count + 1 # 删除最旧的压缩包 for i in range(delete_count): oldest_archive = existing_archives[i] oldest_filename = os.path.basename(oldest_archive) try: # 删除本地文件 os.remove(oldest_archive) print(f"删除最旧的压缩包：{oldest_filename}") # 记录需要清理的文件名 files_to_cleanup.append(oldest_filename) # 删除Hugging Face LFS文件记录 print(f"正在删除Hugging Face LFS文件记录: {oldest_filename}") delete_huggingface_lfs_file(oldest_filename, f"{HF_USER1}/{HF_REPO}", HF_TOKEN1) except Exception as e: print(f"删除失败 {oldest_archive}: {e}") # 如果有文件需要从Git历史中清理，执行安全清理 if files_to_cleanup: print(f"执行安全Git清理: {files_to_cleanup}") safe_git_cleanup(output_dir, files_to_cleanup) # tar.gz 压缩 result = subprocess.run( ['tar', '-czf', output_path, folder_path], capture_output=True, text=True ) if result.returncode == 0: # 计算压缩包大小 file_size = os.path.getsize(output_path) / 1024 / 1024 # 格式化中国时区的时间 china_time = datetime.now(china_tz) formatted_time = china_time.strftime('%Y-%m-%d %H:%M:%S') print(f"压缩成功：{output_path}") print(f"压缩大小：{file_size:.2f} MB") print(f"压缩时间：{formatted_time}") print(f"保留策略：最多保留 {keep_count} 个备份包") # 返回压缩包名和大小信息 return f"{os.path.basename(output_path)} MB:{file_size:.2f}MB TIME:{formatted_time}" else: print("压缩失败") print("错误信息:", result.stderr) return None except Exception as e: print(f"压缩出错: {e}") return None # 调用函数 # new_archive = compress_folder('/data/dv1', 'nezha-hf') def github(type): if type == 1: os.system(f'rm -rf /data/{HF_REPO}') if not os.path.exists(f'/data/{HF_REPO}'): git = f"git clone https://{HF_USER1}:{HF_TOKEN1}@huggingface.co/{HF_USER1}/{HF_REPO}" print(git) os.system(git) os.system(f'git config --global user.email "{HF_EMAIL}"') os.system(f'git config --global user.name "{HF_USER1}"') os.chdir(f'/data/{HF_REPO}') if type == 2: repo_path = f'/data/{HF_REPO}' repo_id = f"{HF_USER1}/{HF_REPO}" os.chdir(f'/data/{HF_REPO}') print("开始备份上传HF") #先清理孤儿LFS引用 cleanup_orphaned_lfs_references(repo_path, repo_id, HF_TOKEN1, keep_count=3) # 执行Git清理（不影响现有文件） os.system('git lfs prune') os.system('git gc --auto') # 备份上传仓库 new_archive_info = compress_folder('/data/dv1', f'/data/{HF_REPO}', keep_count=3) if new_archive_info: new_archive, file_size_info = new_archive_info.split('MB:') os.system(f'git add .') os.system(f'git commit -m "{file_size_info}"') # 正常推送（不再需要强制推送） push_result = os.system('git push origin main') if push_result != 0: print("推送失败，可能有冲突，尝试拉取并合并...") os.system('git pull origin main --rebase') os.system('git push origin main') # 轻度清理 os.system('git gc --auto') os.system('git lfs prune') else: print("压缩失败，无法提交") def nginx(): # 确保目录存在 os.makedirs('/data/nginx1.24', exist_ok=True) # 写入文件 with open('/data/nginx1.24/mime.types', 'w') as f: f.write(mime_types_content) # 设置文件权限（可选） os.chmod('/data/nginx1.24/mime.types', 0o644) print("mime.types 文件已创建") # time.sleep(10) # os.system("rm -rf /data/nginx.conf") # os.system("wget -O '/data/nginx.conf' -q 'https://raw.githubusercontent.com/qilan28/hf-nezha/refs/heads/main/nginx.conf'") with open('/data/nginx.conf', 'w') as f: f.write(nginx_conf) os.system("/data/nginx1.24/sbin/nginx -c /data/nginx.conf") def dv1(): os.system("rm -rf /data/dv1.zip /data/dashboard-linux-amd64 /data/dv1 /data/data") latest_package = get_latest_local_package(f'/data/{HF_REPO}') if latest_package: print(f"最新压缩包路径: {latest_package}") print("通过备份包启动") # tar -xzvf /data/nezha-hf/1759393184994.tar.gz -C /data # tar -xzvf /data/nezha-hf/1759393184994.tar.gz --strip-components=2 -C /data dv1/ print(f"解压：tar -xzvf {latest_package} -C /data") os.system(f"tar -xzvf {latest_package} -C /data") os.system("mv /data/data/dv1/ /data") os.system("rm -rf /data/data") os.chdir('/data/dv1') # 读取 YAML 文件 with open('/data/dv1/data/config.yaml', 'r') as f: config = yaml.safe_load(f) # 修改 listen_port config['listen_port'] = int(N_PORT) # 如果端口号需要是整数 # 写回文件 with open('/data/dv1/data/config.yaml', 'w') as f: yaml.dump(config, f, default_flow_style=False) else: print("通过下载程序启动") if not os.path.exists('/data/dv1'): os.makedirs('/data/dv1') if not os.path.exists('/data/dv1/data'): os.system("rm -rf /data/dv1/data/config.yaml /data/dv1/data/sqlite.db") os.makedirs('/data/dv1/data') with open('/data/dv1/data/config.yaml', 'w') as file: yaml.dump(dashboard_config, file, default_flow_style=False) print("配置文件已写入 /data/dv1/data/config.yaml") print("下载'https://github.com/qilan28/hf-nezha/raw/refs/heads/main/sqlite.db'") os.system("wget -O '/data/dv1/data/sqlite.db' 'https://github.com/qilan28/hf-nezha/raw/refs/heads/main/sqlite.db'") os.chdir('/data/dv1') print(f"下载'https://github.com/nezhahq/nezha/releases/download/{DASHBOARD_VERSION}/dashboard-linux-amd64.zip'") os.system(f"wget -O '/data/dv1/dv1.zip' -q 'https://github.com/nezhahq/nezha/releases/download/{DASHBOARD_VERSION}/dashboard-linux-amd64.zip'") os.system("unzip -o /data/dv1/dv1.zip -d /data/dv1") os.system("rm -rf /data/dv1/dv1.zip") os.system("chmod +x /data/dv1/dashboard-linux-amd64") os.system("mv /data/dv1/dashboard-linux-amd64 /data/dv1/dv1") if os.path.exists('/data/dv1/dv1') and os.path.isfile('/data/dv1/dv1'): print("dv1存在开始启动") threading.Thread(target=repeat_task, daemon=True).start() threading.Thread(target=nginx, daemon=True).start() threading.Thread(target=cf, daemon=True).start() threading.Thread(target=nv1_agent, daemon=True).start() threading.Thread(target=check_system_resources, daemon=True).start() # os.system('/data/dv1/dv1') os.system('nohup /data/dv1/dv1 >> /dev/null 2>&1 &') else: print("dv1不存在") def nv1_agent(): # time.sleep(10) os.system("rm -rf /data/nv1.zip /data/nezha-agent /data/nv1") print(f"下载'https://github.com/nezhahq/agent/releases/download/{NZV1_VERSION}/nezha-agent_linux_amd64.zip'") os.system(f"wget -O '/data/nv1.zip' -q 'https://github.com/nezhahq/agent/releases/download/{NZV1_VERSION}/nezha-agent_linux_amd64.zip'") time.sleep(2) os.system("unzip -o /data/nv1.zip -d /data") os.system("chmod +x /data/nezha-agent") os.makedirs('/data', exist_ok=True) # 写入 YAML 文件 with open('/data/config.yml', 'w') as file: yaml.dump(agent_config, file, default_flow_style=False) print("配置文件已写入 /data/config.yml") time.sleep(2) os.system("rm -rf /data/nv1.zip") os.system("mv /data/nezha-agent /data/nv1") os.system("nohup /data/nv1 -c /data/config.yml >> /dev/null 2>&1 &") def cf(): os.system("rm -rf /data/cf") os.system("wget -O '/data/cf' -q 'https://huggingface.co/datasets/Qilan2/ff/resolve/main/nv1/cf-linux-amd64?download=true'") os.system("chmod +x /data/cf") os.system(f'nohup /data/cf tunnel run --protocol http2 --token {ARGO_AUTH} >> /dev/null 2>&1 &') def _reconstruct_token(partial_token): return partial_token.replace(" ", "") def restart_huggingface_space(space_name, space_id, partial_token): token = _reconstruct_token(partial_token) url = f"https://huggingface.co/api/spaces/{space_name}/{space_id}/restart?factory=true" headers = { "Content-Type": "application/json", "Authorization": f"Bearer {token}", "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/137.0.0.0 Safari/537.36" } try: response = requests.post(url, headers=headers, json={}) return { "status_code": response.status_code, "success": response.status_code == 200, "message": response.text } except requests.RequestException as e: return { "status_code": None, "success": False, "message": str(e) } def check_system_resources(): time.sleep(120) cpu_usage = psutil.cpu_percent(interval=1) memory = psutil.virtual_memory() memory_usage = memory.percent # if cpu_usage >= 90: if cpu_usage >= 90 or memory_usage >= 95: print("占用过高") print(HF_USER2, HF_ID, HF_TOKEN2) result = restart_huggingface_space(HF_USER2, HF_ID, HF_TOKEN2) print(result) else: print("系统资源正常") def repeat_task(): print('备份线程启动') while True: print(f"打包时间：{BACKUP_TIME} 秒") time.sleep(int(BACKUP_TIME))# 2小时 github(2) github(1) os.chdir('/data/') dv1() if os.path.exists('/data/dv1/dv1') and os.path.isfile('/data/dv1/dv1'): print('等待重启中。。。') while True: print(f"打包时间：{RESTART_TIME} 秒") time.sleep(int(RESTART_TIME))# 4小时 14400 6小时 21600 github(2) print(HF_USER2, HF_ID, HF_TOKEN2) result = restart_huggingface_space(HF_USER2, HF_ID, HF_TOKEN2) print(result) # break # github(2) # nv1_agent()