How to Limit CPU Resources in Docker Compose

When running containers in production, proper CPU resource management is crucial for maintaining stable performance. Here's a comprehensive guide on how to limit CPU resources in Docker Compose.

Complete Example

Here's a Docker Compose configuration demonstrating various CPU resource management options:

services:
  api:
    image: node:alpine
    deploy:
      resources:
        limits:
          cpus: '0.50'          # Container can use at most 50% of a CPU
          memory: '512M'        # Maximum memory limit
        reservations:
          cpus: '0.25'          # Guaranteed CPU resources
          memory: '256M'        # Minimum memory reservation
    cpu_shares: 1024            # Relative CPU weight (default: 1024)
    cpuset: '0,1'              # Restrict to CPU cores 0 and 1

  worker:
    image: python:alpine
    deploy:
      resources:
        limits:
          cpus: '0.75'
    cpu_shares: 512             # Half the priority of api service

  cache:
    image: redis:alpine
    cpus: '0.3'                # Direct CPU limit (simpler alternative)

Understanding CPU Options

1. cpus
   • Simplest way to limit CPU usage
   • Value of 1.0 means one CPU core
   • Example: cpus: '0.5' limits to 50% of one CPU
2. cpu_shares
   • Sets relative weight between containers
   • Default value is 1024
   • Only affects containers competing for CPU
3. deploy.resources
   • Modern way to specify resource constraints
   • Supports both limits and reservations
   • Used in Swarm mode and recent Compose versions
4. cpuset
   • Pins container to specific CPU cores
   • Useful for performance-critical applications
   • Format: "0,1" or "0-3"

Monitoring

Monitor CPU usage with:

docker stats
# or for specific service
docker-compose top service_name

Best Practices

1. Start with conservative limits (cpus: '0.5')
2. Use cpu_shares for relative prioritization
3. Monitor and adjust based on actual usage
4. Document resource allocation decisions

Common Issues

• Performance Issues: If service is slow, check if CPU limits are too restrictive
• Uneven Distribution: Use cpu_shares to balance load between services
• High CPU Usage: Monitor with docker stats and adjust limits accordingly

For more advanced resource management, consider using Kubernetes, which offers more sophisticated orchestration capabilities.