Monitoring MCP Servers

This guide explains how to monitor the performance and health of your MCP servers using MCP-Cloud's monitoring features.

Overview

MCP-Cloud provides comprehensive monitoring capabilities for your MCP servers, including:

• Real-time metrics visualization
• Log streaming and analysis
• Alerting and notification
• Historical performance data
• Usage analytics

Metrics Dashboard

The metrics dashboard provides a real-time overview of your server's performance.

Key Metrics

Metric	Description	Typical Values	Alert Threshold
CPU Usage	Percentage of CPU resources used	10-60%	>80%
Memory Usage	Memory consumption in MB	Varies by model	>90% of allocated
Request Rate	Requests per minute	Varies by usage	N/A
Response Time	Average response time in ms	200-2000ms	>5000ms
Error Rate	Percentage of failed requests	<1%	>5%
Token Throughput	Tokens processed per second	Varies by model	N/A

Real-time Logs

MCP-Cloud provides real-time log streaming via Server-Sent Events (SSE), allowing you to monitor your server's operations as they happen.

Accessing Logs

1. Navigate to your server's dashboard
2. Select the "Logs" tab
3. View real-time logs or search historical logs

Log Levels

Logs are categorized by severity:

• DEBUG: Detailed information for debugging
• INFO: General operational information
• WARN: Warnings that don't affect functionality
• ERROR: Errors that affect functionality
• FATAL: Critical errors that cause service failure

Log Filtering

You can filter logs by:

• Time range
• Log level
• Message content
• Request ID

SSE Metrics Stream

For programmatic monitoring, MCP-Cloud provides a metrics stream via SSE:

const eventSource = new EventSource(
  'https://api.mcp-cloud.ai/api/servers/123/metrics/stream',
  {
    headers: {
      'Authorization': 'Bearer YOUR_API_TOKEN'
    }
  }
);

eventSource.addEventListener('metrics', (event) => {
  const metrics = JSON.parse(event.data);
  console.log('Current CPU usage:', metrics.cpu, '%');
  console.log('Current memory usage:', metrics.memory, 'MB');
  console.log('Requests in last minute:', metrics.requestRate);
});

Setting Up Alerts

MCP-Cloud allows you to configure alerts based on metrics thresholds:

1. Navigate to your server's dashboard
2. Select the "Alerts" tab
3. Click "Create Alert"
4. Configure alert conditions:
   • Metric: Select the metric to monitor
   • Threshold: Set the trigger threshold
   • Duration: How long the condition must persist
   • Channel: Email, SMS, or webhook notification

Alert Channels

MCP-Cloud supports multiple notification channels:

• Email notifications
• SMS alerts
• Webhook integration
• Slack integration
• PagerDuty integration

Historical Analytics

MCP-Cloud retains performance data for historical analysis:

• 1-minute resolution for the last 24 hours
• 1-hour resolution for the last 30 days
• 1-day resolution for the last year

Access historical data through:

1. The web dashboard
2. CSV export
3. API access

Resource Optimization

Based on monitoring data, MCP-Cloud can recommend resource optimizations:

• Memory allocation adjustments
• CPU allocation adjustments
• Scaling configuration changes
• Region recommendations

Integration with External Monitoring

MCP-Cloud can integrate with external monitoring systems:

• Prometheus metrics export
• Grafana dashboards
• Datadog integration
• New Relic integration

Example Prometheus configuration:

scrape_configs:
  - job_name: 'mcp-server'
    scrape_interval: 15s
    metrics_path: '/metrics'
    scheme: https
    basic_auth:
      username: 'prometheus'
      password: 'YOUR_API_TOKEN'
    static_configs:
      - targets: ['your-server-id.metrics.mcp-cloud.ai']

Best Practices

1. Set up alerts for critical metrics: CPU, memory, error rate
2. Monitor response times: Identify performance degradation early
3. Review logs regularly: Look for warning patterns before they become errors
4. Track usage patterns: Understand peak usage times
5. Correlate metrics with business events: Identify impact of marketing campaigns, etc.
6. Test alert configurations: Ensure notifications work as expected

Troubleshooting

If monitoring data appears incorrect:

1. Check the server status to ensure it's running
2. Verify that any proxy or firewall isn't blocking metrics
3. Ensure your API token has the necessary permissions
4. Try accessing raw metrics via the API to bypass dashboard issues
5. Contact support if issues persist