SSH MCP Troubleshooting Guide

Quick Diagnostics

Check SSH Connection

ping -c 4 43.143.207.242
nc -zv 43.143.207.242 22
telnet 43.143.207.242 22

Check Server Status

systemctl status sshd
systemctl status firewalld
iptables -L -n
df -h
free -m
uptime

Common Issues

1. Connection Timeout

Symptoms:

• "Connection timed out"
• "No route to host"

Solutions:

1. Check server IP is correct
2. Check server is running
3. Check firewall allows SSH (port 22)
4. Check server has network connectivity
5. Try with longer timeout: timeout=120

MCP Example:

连接 SSH，host=43.143.207.242, username=root, password=xxx, timeout=120

2. Authentication Failed

Symptoms:

• "Authentication failed"
• "Permission denied"

Solutions:

1. Verify username is correct
2. Verify password is correct
3. Check SSH config: PasswordAuthentication yes
4. Check user exists on server
5. Check user has shell access: grep username /etc/passwd

Server Commands:

sudo grep "PasswordAuthentication" /etc/ssh/sshd_config
sudo systemctl restart sshd

3. Error Reading SSH Protocol Banner

Symptoms:

• "Error reading SSH protocol banner"
• "Connection closed by remote host"

Solutions:

1. SSH service may be hung - restart it:

   sudo systemctl restart sshd

1. Check SSH service status:

   sudo systemctl status sshd
   sudo journalctl -u sshd -n 50

1. Check server load:

   uptime
   top

1. Check SSH port is responding:

   nc -zv hostname 22

4. Host Key Verification Failed

Symptoms:

• "Host key verification failed"
• "Known hosts file error"

Solutions:

1. Remove old host key:

   ssh-keygen -R hostname

1. Or use accept_new_host_key=true in ssh_connect (testing only):

   连接 SSH，host=xxx, accept_new_host_key=true

1. Or disable strict checking (not recommended for production):

   连接 SSH，host=xxx, strict_host_key_checking=false

5. Connection Refused

Symptoms:

• "Connection refused"
• "No connection could be made"

Solutions:

1. SSH service not running:

   sudo systemctl start sshd
   sudo systemctl enable sshd

1. SSH on different port - check port:

   grep "^Port" /etc/ssh/sshd_config

6. Command Blocked by Security Policy

Symptoms:

• "Command blocked by security policy"
• "SecurityError" in output
• Interactive resolution instructions shown

Solutions:

Method 1: Adjust security level (recommended)

{
  "mcpServers": {
    "ssh": {
      "command": "ssh-licco",
      "env": {
        "SSH_SECURITY_LEVEL": "relaxed"
      }
    }
  }
}

Method 2: Add extra allowed commands

{
  "SSH_EXTRA_ALLOWED_COMMANDS": "your-command"
}

Method 3: Add extra allowed patterns (for special characters)

{
  "SSH_EXTRA_ALLOWED_PATTERNS": "|,>,<,&,;"
}

Security Levels:

7. Rate Limit Triggered

Symptoms:

• "Rate limit exceeded" message
• "超过 30 次请求/60秒"

Solutions:

1. Wait and retry (default: 30 requests per 60 seconds)
2. Adjust limits:

   {
     "SSH_RATE_LIMIT_MAX": "60",
     "SSH_RATE_LIMIT_WINDOW": "60"
   }

1. Disable rate limiting (not recommended for production):

   {
     "SSH_RATE_LIMIT": "false"
   }

8. Docker Build Failed

Symptoms:

• "Docker build failed"
• "Cannot connect to Docker daemon"

Solutions:

1. Check Docker is installed:

   docker --version

1. Check Docker service:

   sudo systemctl status docker
   sudo systemctl start docker

1. Add user to docker group:

   sudo usermod -aG docker $USER

1. Check Docker socket permissions:

   ls -la /var/run/docker.sock

9. File Transfer Failed

Symptoms:

• "SFTP error"
• "Permission denied"

Solutions:

1. Check remote directory exists:

   ls -la /remote/path/

1. Check write permissions:

   touch /remote/path/test.txt

1. Use correct path separators (Linux uses / not \)

10. Background Task Not Found

Symptoms:

• "NOT_FOUND" status when checking task
• PID file missing

Solutions:

1. Check if the task command was correct
2. Verify the task_id matches
3. Check if the session is still active:

   列出 SSH 会话

1. Check logs directly:

   执行命令，command=cat /tmp/background_task.log, session_id=xxx

11. Password Conflict Detected

Symptoms:

• ssh_host shows password conflict warning
• Different passwords in MCP config vs hosts.json

Solutions:

1. Unify passwords across config sources
2. Or use force env mode:

   {
     "SSH_FORCE_ENV_CONFIG": "true"
   }

1. This ensures MCP env config always takes priority

12. npx Cannot find module Error (Startup Failure)

Symptoms:

• 配置 "command": "npx", "args": ["ssh-licco"] 后 MCP 客户端报错
• 错误包含 Cannot find module '.../node_modules/ssh-licco/bin/ssh-licco.js'
• 错误包含 MODULE_NOT_FOUND

Cause: npm 全局目录中存在损坏/不完整的 ssh-licco 包，导致 npx 加载本地缓存而非下载新包。

Solution 1: Clean npm global package (recommended):

npm uninstall -g ssh-licco

卸载后 npx ssh-licco 会重新从 npm registry 下载完整包，自动安装 Python 依赖并运行。

Solution 2: Use local source directly (for development): 如果本地有项目源码，将 MCP 配置改为直接引用本地文件：

{
  "mcpServers": {
    "ssh-licco": {
      "command": "node",
      "args": ["D:\\path\\to\\ssh-licco\\ssh-licco.js"],
      "env": {
        "SSH_LICCO_AUTO_INSTALL": "true",
        ...
      }
    }
  }
}

Solution 3: Run smart installer directly:

python smart_install.py

Prevention:

• 初次使用 npx ssh-licco 前，确保没有全局安装的损坏包
• 如果全局安装过旧版本，先 npm uninstall -g ssh-licco 清理
• 避免将全局 npm 缓存路径设置为项目源码路径

13. Anaconda ModuleNotFoundError: No module named 'ssh_mcp'

Symptoms:

• 错误栈指向 D:\software\anaconda\Scripts\ssh-licco.exe\__main__.py
• ModuleNotFoundError: No module named 'ssh_mcp'
• pip show ssh-licco 显示 WARNING: Ignoring invalid distribution ~sh-licco
• where ssh-licco 找不到命令，但 exe 文件实际存在

Cause: Anaconda 的 site-packages 目录中存在带 ~ 前缀的损坏残留目录（如 ~sh_licco-0.5.5.dist-info），pip 的 console_scripts 入口点（ssh-licco.exe）存在但模块已丢失。通常由 pip 安装过程中断导致。

Diagnosis Flow（复现该问题的完整排查步骤）：

# 1. 检查 pip 中包的状态
pip show ssh-licco
# → WARNING: Ignoring invalid distribution ~sh-licco
# → Package(s) not found: ssh-licco

# 2. 检查可执行文件
where ssh-licco
# → 可能找不到（PATH 未包含 Anaconda Scripts）
Get-ChildItem D:\software\anaconda\Scripts\ssh-licco*
# → ssh-licco.exe 存在（入口点残留）

# 3. 检查 site-packages 中的残留
Get-ChildItem D:\software\anaconda\Lib\site-packages | Where-Object { $_.Name -like "~*licco*" }
# → ~sh_licco-0.5.5.dist-info 目录（带 ~ 前缀 = pip 标记为损坏）

# 4. 尝试直接导入验证
D:\software\anaconda\python.exe -c "from ssh_mcp.cli import main; print('OK')"
# → ModuleNotFoundError: No module named 'ssh_mcp'

# 5. 尝试 pip 安装后再次验证
D:\software\anaconda\python.exe -m pip install --force-reinstall ssh-licco
# → 成功安装 0.5.4 及 33 个依赖
# → 仍报 WARNING: Ignoring invalid distribution ~sh-licco（~ 目录未清理）

Fix: Diagnosis-based repair:

# Step 1: Clean corrupted ~-prefixed dist-info using Python（shell 权限限制时用）
D:\software\anaconda\python.exe -c "import shutil, os; path = r'D:\software\anaconda\Lib\site-packages'; [shutil.rmtree(os.path.join(path, d), ignore_errors=True) for d in os.listdir(path) if d.startswith('~') and 'licco' in d.lower()]"

# Step 2: Verify cleanup
D:\software\anaconda\python.exe -m pip show ssh-licco
# → WARNING 消失
# → Version: 0.5.4 (from PyPI)

# Step 3: Install local dev version (editable) to get latest code
D:\software\anaconda\python.exe -m pip install -e D:\pyworkplace\ssh-mcp
# → Successfully installed ssh-licco-0.5.5

# Step 4: Verify complete import chain
D:\software\anaconda\python.exe -c "from ssh_mcp.server import SSHMCPServer; from ssh_mcp.session_manager import SessionManager; from ssh_mcp.cli import main; print('OK')"
# → OK ✅

Key Commands Quick Reference:

Prevention:

• 如果要在 Anaconda 中使用，统一用 D:\software\anaconda\python.exe -m pip install 操作
• 不要在安装过程中中断 pip（可能导致 ~ 前缀的残留）
• ssh-licco 的 npx 方式使用独立 venv（~/.ssh-licco-venv），不受 Anaconda 环境影响
• 排查同类问题时，先走 pip show → Get-ChildItem ~* → python.exe -c "import" 三步诊断

Diagnostic Commands

Server Diagnostics

uname -a
cat /etc/os-release
ip addr
ip route
ss -tuln
df -h
free -h
uptime
top -bn1 | head -15
ps aux | head -20
ps -ef | grep ssh
tail -f /var/log/secure
journalctl -u sshd -n 100

MCP Diagnostics

import json
password = "P/[KY}+wa7?2|uc"
print(json.dumps({"password": password}))

Security Diagnostics

echo $SSH_SECURITY_LEVEL
echo $SSH_RATE_LIMIT
echo $SSH_EXTRA_ALLOWED_COMMANDS

Health Check Script

Run on remote server:

#!/bin/bash
echo "=== System Info ==="
uptime
echo ""
echo "=== Disk Usage ==="
df -h
echo ""
echo "=== Memory ==="
free -h
echo ""
echo "=== SSH Service ==="
systemctl status sshd --no-pager
echo ""
echo "=== Network Connections ==="
ss -tuln | grep :22
echo ""
echo "=== Last Login ==="
last -5

Reset SSH Service (Server Side)

sudo systemctl stop sshd
pkill -9 sshd
rm -rf /run/sshd*
sudo systemctl start sshd
sudo systemctl status sshd

Log Locations

Get Help

Collect Debug Info

script debug_session.log
uname -a
df -h
free -h
systemctl status sshd
netstat -tuln | grep 22
echo "Security Level: $SSH_SECURITY_LEVEL"
echo "Rate Limit: $SSH_RATE_LIMIT"
exit

Report Issue

Include:

1. Error message
2. Server IP/hostname
3. OS version: cat /etc/os-release
4. SSH version: ssh -V
5. SSH_SECURITY_LEVEL setting
6. Relevant log excerpts

Common Server Fixes

Restart SSH Service

sudo systemctl restart sshd

Check SSH Config

sudo sshd -t
sudo systemctl reload sshd

Fix Permissions

chmod 700 ~/.ssh
chmod 600 ~/.ssh/authorized_keys
chmod 644 ~/.ssh/known_hosts

Check User Shell

grep username /etc/passwd

Allow Root Login (if needed)

sudo nano /etc/ssh/sshd_config
PermitRootLogin yes
PasswordAuthentication yes
sudo systemctl restart sshd