Massed Compute

When you’re debugging a GPU cloud deployment or troubleshooting workload issues, you need visibility into what’s actually running on your VMs. This guide shows you how to generate structured health and workload inventory reports for existing Massed Compute instances without disrupting running services or reinstalling software stacks.

The inventory process captures SSH connectivity, GPU visibility, disk usage hotspots, active services, listening ports, Python environments, and system resources in a single structured report that both humans and AI agents can analyze for debugging.

Step-by-Step VM Health Inventory

# List your running instances
curl -H "Authorization: Bearer MC_TOKEN" \
     https://vm.massedcompute.com/api/instances

# Note the UUID of your target instance
export INSTANCE_UUID="your-instance-uuid-here"

# Test SSH connection
ssh -i ~/.ssh/your-key.pem ubuntu@YOUR_VM_IP \
    -o BatchMode=yes \
    -o ConnectTimeout=10 \
    'echo "SSH connectivity: OK"'

# Clear any stale host keys if needed
ssh-keygen -R YOUR_VM_IP

# System overview
ssh -i ~/.ssh/your-key.pem ubuntu@YOUR_VM_IP 'bash -s' << 'EOF'
echo "=== System Info ==="
hostname && uptime && whoami
echo

echo "=== Disk Usage ==="
df -h
echo

echo "=== Memory Usage ==="
free -h
echo

echo "=== CPU Info ==="
nproc && lscpu | grep "Model name"
EOF

# GPU inventory
ssh -i ~/.ssh/your-key.pem ubuntu@YOUR_VM_IP 'bash -s' << 'EOF'
echo "=== GPU Status ==="
if command -v nvidia-smi &> /dev/null; then
    nvidia-smi --query-gpu=name,memory.total,memory.used,temperature.gpu --format=csv,noheader,nounits
    echo
    nvidia-smi -L
else
    echo "No NVIDIA GPUs detected or drivers not installed"
fi
EOF

# Services and ports inventory
ssh -i ~/.ssh/your-key.pem ubuntu@YOUR_VM_IP 'bash -s' << 'EOF'
echo "=== Active Services ==="
systemctl list-units --type=service --state=active --no-pager | head -20
echo

echo "=== Failed Services ==="
systemctl --failed --no-pager
echo

echo "=== Listening Ports ==="
ss -tlnp | grep LISTEN
echo

echo "=== Top Processes ==="
ps aux --sort=-%mem | head -10
EOF

# Python environment inventory
ssh -i ~/.ssh/your-key.pem ubuntu@YOUR_VM_IP 'bash -s' << 'EOF'
echo "=== Python Environments ==="
if [ -n "$VIRTUAL_ENV" ]; then
    echo "Active venv: $VIRTUAL_ENV"
else
    echo "No active virtual environment"
fi
echo

echo "=== Python Versions ==="
python3 --version 2>/dev/null || echo "Python3 not found"
pip --version 2>/dev/null || echo "Pip not found"
echo

echo "=== Conda Environments ==="
if command -v conda &> /dev/null; then
    conda env list
else
    echo "Conda not installed"
fi
EOF

# Network connectivity test
ssh -i ~/.ssh/your-key.pem ubuntu@YOUR_VM_IP 'bash -s' << 'EOF'
echo "=== Network Connectivity ==="
curl -s -o /dev/null -w "Hugging Face: %{http_code} (%{time_total}s)\n" https://huggingface.co || echo "Hugging Face: Failed"
curl -s -o /dev/null -w "PyPI: %{http_code} (%{time_total}s)\n" https://pypi.org || echo "PyPI: Failed"
echo

echo "VM_HEALTH_INVENTORY_OK"
EOF

Troubleshooting Common Issues

SSH Connection Problems

If SSH fails, check that your instance is running and the security group allows SSH access on port 22.

# Check instance status via API
curl -H "Authorization: Bearer MC_TOKEN" \
     "https://vm.massedcompute.com/api/instances/$INSTANCE_UUID"

# Remove stale host keys
ssh-keygen -R YOUR_VM_IP

GPU Not Detected

If nvidia-smi fails on a GPU instance, you may have launched a CPU SKU by mistake or need to install drivers.

# Check if this is actually a GPU instance
lspci | grep -i nvidia

# If GPUs are present but nvidia-smi fails, reinstall drivers
sudo apt update && sudo apt install -y nvidia-driver-535

High Disk Usage

Use the disk usage output to identify space-consuming directories and clean up as needed.

# Find largest directories
du -h --max-depth=2 ~ | sort -hr | head -10

# Clean common temporary locations
rm -rf ~/.cache/pip/*
rm -rf /tmp/*

Skip All of This: Deploy with an AI Agent

This entire health inventory process exists as a tested, machine-readable recipe in the Massed Compute MCP. Instead of running SSH commands manually, you can connect an AI agent that handles the VM identification, connectivity testing, and inventory generation automatically.

{
  "mcpServers": {
    "massed-compute": {
      "type": "http",
      "url": "https://vm.massedcompute.com/api/mcp",
      "headers": { "Authorization": "Bearer MC_TOKEN" }
    }
  }
}

Then say:

The agent matches this request against the recipe catalog (tested June 2026), identifies your running VM, establishes SSH connectivity, and executes the complete inventory workflow above. It provisions the right connection parameters, runs all health checks in the correct sequence, and reports back with structured results. If any connectivity or inventory step fails, the agent stops and reports the exact error for debugging.

Quick Setup Guide

For immediate VM health monitoring:

1. Launch instance: Use any Massed Compute GPU or CPU VM with Ubuntu 24.04
2. Test SSH: Verify connectivity with your private key
3. Run inventory: Execute the remote SSH commands from steps 1-7
4. Analyze output: Review system resources, GPU status, and service health
5. Debug issues: Use troubleshooting section for common problems

Frequently Asked Questions