Commit Graph

26 Commits

Author SHA1 Message Date
Nexus Agent 84b388fae2 fix(sync): rsync 推送非 root 自动 sudo 提权
文件推送与文件管理对齐:非 root 默认 auto_sudo 时使用 sudo -n rsync;
新增批量 sudoers 补丁脚本以补全 rsync 白名单。
2026-06-11 23:20:45 +08:00
Nexus Agent c8b0663508 feat: 批量IP添加、脚本库历史、推送权限记录与脚本SSH执行修复
Nexus CI/CD / test (push) Waiting to run
Nexus CI/CD / deploy (push) Blocked by required conditions
Nexus Pre-commit Checks / quick-check (push) Waiting to run
批量添加替代CSV并支持去重;脚本库页内嵌执行历史;推送历史记录 rsync 权限策略;
脚本执行不再依赖 Agent 在线;服务器同步日志与相关单测补齐。

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-08 03:15:40 +08:00
Nexus Agent f12a0b6a36 fix(sync): rsync 推送后设置远程属主 www 与权限 755
推送默认 --chown=www:www --chmod=D755,F755,避免文件落盘为 root。

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-08 02:23:43 +08:00
Nexus Agent 5ed3c28491 fix(sync): 推送异常时落库 failed,修复订正时区与卡住阈值
rsync 抛错时 sync_log 不再永久 running;reconcile 兼容 naive UTC。

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-08 02:14:33 +08:00
Nexus Deploy b866e944ab fix: Code Review P0/P1 batches 1-3 and single-operator risk acceptance
Apply sync/install/auth/schedule/retry/agent/settings fixes from full
code review; document accepted WS and Agent legacy risks for solo ops.

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-04 15:57:49 +08:00
Nexus Deploy a2ae74d582 release: BUG fixes batch 1-6, full bug scan docs, Ubuntu/Linux dev
- Backend: auth refresh reuse, schedule/retry/sync/agent/files fixes
- Frontend: push WS, files ETag browse, vite build assets
- Docs: audit/changelog, production deploy gate, bug scan registry B7-77
- Deploy: deploy-production.sh, prune assets, gate logs
2026-06-04 14:01:14 +08:00
Your Name f078d0e03b feat(push): localize sync/rsync error messages to Chinese in history and progress
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-02 15:18:01 +08:00
Your Name 78798e3630 feat(push): refactor B0-B6, /ws/sync channel, staging cleanup
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-02 08:14:09 +08:00
Your Name 832e34fd98 fix: 修复 6 个 MEDIUM 级 Bug (锁定/重试/调度/分页/SSH池/token_version)
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-02 01:29:06 +08:00
Your Name 19cfb7caaa fix: 系统全扫描修复 7 个 HIGH 级 Bug
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-02 01:22:00 +08:00
Your Name d93c518a14 feat(files): POSIX paths, elevation, recursive chmod, access hints and docs
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-01 19:54:18 +08:00
Your Name 32feb1b6db fix: 全站 ruff 清零 — 77 errors → 0
Fixes:
- F821: install.py site_url 未定义(NameError)、sync_v2.py os 未导入、script_jobs.py LOG f-string
- F401: 移除 22 个未使用导入(models/__init__.py、install.py、auth.py 等)
- B904: 20 个 except 块 raise 添加 from e/from None
- S110: 20 个有意的 try-except-pass 添加 noqa 注释(SSH清理/DDL幂等/WS断连)
- B007: 3 个未使用循环变量重命名(dirs→_dirs、server_id→_server_id)
- F541: 3 个无占位符 f-string 修正
- F841: auth.py 未使用 ip_address 变量移除
- S105: auth_service.py Redis key prefix 误报 noqa
- B023: script_execution_flush.py 循环变量绑定修复

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-05-30 20:07:45 +08:00
Your Name 0b82a3fee7 feat: 推送取消 — 引擎 cancel flag 检查 + WS cancelled 计数器
sync_engine_v2: 每台服务器推送前检查 Redis sync:cancel:{batch_id},
若已取消则跳过并标记 cancelled 状态,WS 广播 cancelled 计数。
websocket: broadcast_sync_progress 新增 cancelled 参数。

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-30 00:57:38 +08:00
Your Name 74de9d303e feat: 推送页面 Round 2 迭代 — 同步说明 + 在线标识 + 失败重试 + Telegram通知 + 文件预览
F5: 同步模式详细说明 — 每种模式下方显示说明文字,切换时动态更新
F2: 服务器在线状态标识 — 服务器列表显示🟢/🔴在线/离线标识
F1: 失败自动重试 — 推送失败自动创建重试任务 + 失败行"🔄 重试"按钮
F3: 推送完成 Telegram 通知 — 全成功🟢/部分失败🟡/全失败🔴 三种消息含详情
F4: 文件内容预览 — 文件管理器点击文件名弹出内容预览模态框

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-29 15:59:30 +08:00
Your Name 82c297776a feat: 推送页面 5 项迭代 — WS实时进度 + 分组选择 + 历史增强 + 文件操作 + 推送校验
F1: broadcast_sync_progress() + batch_id + 前端WS逐台更新
F2: get_paginated() platform_id/node_id + /categories端点 + 筛选下拉框
F4: _sync_log_to_dict() 补全字段 + diff_summary捕获 + 可展开详情面板
F7: /local-file-ops端点 + 文件管理器删除/重命名
F8: /verify端点 md5sum对比 + 推送后校验UI

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-28 21:01:27 +08:00
Your Name 97be1fd214 完善审计日志中文化:auth_service/script_service/sync_engine 全部 detail 改中文 + 前端新增 webssh_token/refresh_token_mismatch 等映射 + MCP deploy 修复
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-27 02:57:51 +08:00
Your Name cdd0be328a fix: 六轮深度扫描 — 47项Bug修复、安全加固、死代码清理
Critical runtime bugs:
- terminal.html WebSSH完全不可用(URL前缀/JSON解析/Content-Type三处错误)
- servers.py路由遮蔽:/logs被/{id}拦截,3个前端页面同步日志查询失败
- scripts.html startExecPoll()→startExecPolling(),长任务快速执行崩溃
- agent.py {value!r!s:.50}格式串非法,agent发非数值时ValueError
- alerts.html d.daily.reduce()无null检查,API返回空数据时TypeError

Resource leak / stability:
- websocket.py僵尸连接未关闭TCP,文件描述符泄漏
- websocket.py _last_alert_time字典无限增长(加1小时过期清理)
- asyncssh_pool.py全忙时超过MAX_CONNECTIONS无限增长
- self_monitor.py Telegram告警无冷却,宕机时每30秒刷屏
- schedule_runner.py一次性调度执行超60秒会重复触发
- 限速脚本EXPIRE每次重置窗口可绕过(改用Lua原子脚本)

Security:
- JWT access token加token_version声明,改密码后旧token立失效(零宽限)
- INSTALL_MODE导入时常量→动态函数,安装后JWT认证不再残留禁用
- install.py /lock端点加管理员存在性验证,防止阻断安装
- ServerUpdate schema移除connectivity只读字段,防止伪造连接状态

Frontend fixes:
- doExec()缺r.ok检查、commands.html null检查
- _server_to_dict()补last_checked_at+ssh_key_public
- _field_match()逗号cron表达式修复
- alerts类型显示、SSH会话名称、搜索高亮定位
- 一次性/循环定时任务(run_mode+fire_at+自动禁用)

Dead code removed (400+ lines):
- SyncService batch_push/_push_single等5个方法(零调用者)
- 5个未使用schema(SyncCommands/SyncConfig/SyncSftp/FileDeploy/PaginatedResponse)
- 6个零调用service方法、3个无前端API端点
- 4个未使用import

Schema migrations:
- push_schedules: run_mode + fire_at列,cron_expr改NULL
- servers: 7个新列 + ssh_key_private/public VARCHAR(500)→TEXT

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2026-05-24 16:26:40 +08:00
Your Name ac0eb79707 feat: rsync dry-run preview (方案D 智能提示型)
Backend:
- schemas.py: SyncPreview model (server_id, source/target_path, sync_mode, verbose)
- sync_engine_v2.py: preview_sync() + _parse_rsync_stats() helpers
  - runs rsync --dry-run --stats; verbose=True adds -v (capped 200 lines)
  - returns stats dict: files_transferred/created/deleted + byte counts
- sync_v2.py: POST /api/sync/preview endpoint with audit log

Frontend (push.html - 方案D):
- Sync mode radio change listener: onSyncModeChange()
- full/overwrite mode: orange warning card + prominent preview button
- incremental/checksum mode: subtle "预览首台变动 →" link
- Inline preview result section (no modal): stats cards + optional file list
- files_deleted shown in red card for destructive modes
- verbose toggle: "显示详细文件列表 ▼" triggers second request with verbose=true
- fmtBytes() helper for human-readable file sizes

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-23 16:41:47 +08:00
Your Name c3c3bae370 fix: Phase 3g sync_engine_v2 bugs
F3g-01 sync_commands line 195: server_results is a list; unpacking it
       with ** raises TypeError, silently caught by gather(return_exceptions=True),
       so S2 command sync and S3 config sync returned empty results for all
       servers. Fixed by using a proper dict: {"server_name": ..., "results": ...}.

F3g-02 sync_config: config value containing newline chars was not stripped
       before shlex.quote; echo inside single-quotes would write multiple lines
       to the sysctl config file, injecting arbitrary keys. Strip \n\r\x00 from
       value before quoting (requires admin auth, defense-in-depth fix).

scripts.py, settings.py: CLEAN (no changes needed).

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-23 15:46:52 +08:00
Your Name 752a24497c feat: Phase 2 security audit fixes + script execution platform
Security (P0/P1/P2):
- WebSSH: dedicated short-lived token, server_id binding, 4003 close code
- Remove /api/agent/exec from control plane (RCE surface eliminated)
- Global JWT middleware (JwtAuthMiddleware) with install-mode bypass
- decrypt_value: failure returns None, never leaks ciphertext
- Telegram: sanitize_external_message strips sensitive fields
- Agent auth: startup enforces non-empty API key, compare_digest
- Credentials: password_set bool flag, no plaintext in API responses
- audit_log: writes admin_username + ip_address on all CUD ops
- Install lock: all /api/install/* except GET /status return 403 post-install
- WebSocket dedup: publish once, subscribers deliver locally

Script execution platform:
- script_jobs.py / script_job_callback.py: async batching and callback
- script_execution_store.py / script_callback_rate.py: Redis-backed state
- script_execution_flush.py: background flusher to MySQL
- scripts.html / script-executions.html: full execution UI
- agent_url.py: centralised URL builder

Frontend:
- All 13 pages migrated from CDN to /app/vendor/ (no external deps)
- vendor/: alpinejs, tailwindcss-browser, xterm, xterm-addon-*, qrious
- Dashboard WebSocket real-time alerts; 8h JWT session timeout

Tests:
- test_security_unit.py: 15 unit tests (JWT, sanitize, vendor, install 403)
- test_api.py: env-configurable admin credentials

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-23 15:26:56 +08:00
Your Name d744d7df8e 安全与质量修复: P0-4/P0-5 + P1-6~P1-13 (8项)
P0-4: install.py _write_env() 值加引号转义,防止含#$\等字符的密码破坏.env格式
P0-5: install.py _build_redis_url() URL编码redis密码,防止@:等字符破坏URL解析
P1-6: auth_service.py _decode_access_token() 验证exp/sub声明存在,拒绝畸形JWT
P1-7: websocket.py + webssh.py WebSocket JWT验证增加updated_at检查,密码修改后令牌失效
P1-8: auth.py 无refresh_token的logout返回更明确的提示信息
P1-9: install.py create_admin增加_is_installed()检查,防止.env存在后重复创建
P1-10: servers.py server_stats() 改用SQL聚合查询,避免加载全部服务器对象
P1-11: heartbeat_flush.py 移除get_redis()永不返回None的死代码
P1-13: sync_engine_v2.py completed/failed计数器加asyncio.Lock防止并发竞态

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-22 23:04:40 +08:00
Your Name 938d26927f P1/P2: 后端安全与稳定性修复
- Agent per-server API Key认证 (agent.py)
- JWT updated_at校验与会话超时 (auth_jwt.py)
- 服务器凭据Fernet加密存储+密码脱敏 (servers.py)
- WebSSH服务器级授权检查 (webssh.py)
- Config同步去重+回滚机制 (sync_engine_v2.py)
- DB层分页offset/limit (server_repo.py)
- session.py logger修复 + @property死代码清理
- 心跳刷入rollback误用修复 (heartbeat_flush.py)

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-22 22:28:57 +08:00
Your Name c9a99f4fb3 fix: 全项目文档对齐 + 代码清理
代码修复:
- web/agent/agent.py: datetime.utcnow() → datetime.now(timezone.utc)
- requirements.txt: 移除 paramiko==3.5.0 (已无活跃引用)
- 删除 server/infrastructure/ssh/pool.py (DEPRECATED, 无引用)

连接池三层对齐 (config.py/.env.example/session.py):
- DB_POOL_SIZE 100→160, DB_MAX_OVERFLOW 100→120
- 基于 MySQL max_connections=400, install.php 公式

文档修复 (21项):
- P0: 硬编码域名/IP替换为配置变量占位符
- P1: 10个过时设计文档加归档标注 (引用旧文件结构)
- P2: step-3-webssh报告 paramiko引用修正
- 6份审查报告连接池参数勘误 100/100→160/120
- ECC安全报告 EC5 datetime.utcnow 标记已修复
- docs/README.md 文档索引重写
- docs/memory/mem_nexus_overview.md 移除硬编码凭证
- docs/project/tech-stack-inventory.md paramiko标记已移除

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-22 08:19:56 +08:00
Your Name 5590391779 feat: Nexus Agent rebrand + server_id + exec endpoint + trigger_type
Nexus CI/CD / test (push) Waiting to run
Nexus CI/CD / deploy (push) Blocked by required conditions
Nexus Pre-commit Checks / quick-check (push) Waiting to run
Agent (agent.py):
- Rebrand MultiSync → Nexus throughout
- Add server_id to heartbeat payload (required by backend AgentHeartbeat schema)
- Add /exec endpoint for remote command execution via Nexus
- Update heartbeat fields to match backend: cpu_usage, mem_usage, disk_usage
- Use 60s default interval (matches CLAUDE.md spec)

Agent (agent.sh):
- Rebrand, add SERVER_ID env var, send in heartbeat payload
- Update service name to nexus-agent

Agent (install.sh):
- Rebrand, add --id parameter for server_id
- Write server_id to config.json
- Service name: nexus-agent
- Add config.example.json for reference

SyncEngineV2:
- Add trigger_type parameter to sync_files() (was hardcoded "manual")
- Schedule/retry runners can now pass "schedule"/"retry" trigger types

Nginx:
- Add /agent/ to static asset locations for agent file downloads

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-22 00:37:39 +08:00
Your Name 73c1776e1e fix: P0 critical runtime issues
1. conftest.py: add missing `from server.main import app` import
2. dependencies.py: get_db() reuses middleware session instead of
   creating independent AsyncSessionLocal (fixes double-session bug)
3. auth_jwt.py: get_optional_admin uses request.state.db instead of
   creating leaked AsyncSessionLocal session
4. config.py: default DATABASE_URL uses mysql+aiomysql (not pymysql)
5. sync_engine_v2.py: sanitize config keys with regex, escape values
   to prevent command injection in sync_config
6. sync_v2.py: add POST /api/sync/browse endpoint for remote dir listing
7. files.html: browseDir calls real API instead of placeholder stub

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-21 22:54:35 +08:00
Your Name 539c33ef42 feat: Nexus 6.0 6-step implementation (Step 0-5)
Step 0: Infrastructure hardening — Redis BlockingConnectionPool,
WebSocket two-layer (memory+Redis Pub/Sub), JWT auth, DB session leak fix
Step 1: Data layer — 4 new tables (platforms/nodes/command_logs/ssh_sessions),
data migration (category→Node), repos + API routes
Step 2: Auth layer — JWT middleware on all routes, TOTP JWT integration
Step 3: Web SSH — asyncssh connection pool, /ws/terminal endpoint,
xterm.js frontend, Koko protocol
Step 4: Sync engine v2 — file/command/config/SFTP modes, parallel execution
Step 5: Frontend migration — 12 Tailwind CSS v4 + Alpine.js pages,
PHP-FPM removal from nginx config

21 Python backend + 12 HTML frontend + 2 deploy configs + 3 test files

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-21 22:11:38 +08:00