Docker containers, when left unconstrained, can consume excessive host memory and degrade system performance. Implementing resource limits and periodic maintenance procedures ensures stable cluster operations.

Resource Constraints

Apply hard memory limits during container initialization to prevent individual workloads from exhausting host resources. The following example restricts a container to 512 megabytes of RAM with a 1 gigabyte swap ceiling:

docker run -d \
  --name api-gateway \
  --memory="512m" \
  --memory-swap="1g" \
  --memory-reservation="256m" \
  nginx:alpine

The --memory-reservation flag sets a soft limit that the container may exceed temporarily during traffic spikes, while --memory enforces an absolute cap.

Monitoring Allocation

Track real-time consumption across all running instances using formatted output for better readability:

docker stats --all \
  --format "table {{.Container}}	{{.Name}}	{{.MemUsage}}	{{.NetIO}}"

For automated alerting, parse the JSON stream programmatically:

docker stats --no-stream --format "{{json .}}" | \
  jq -r 'select(.MemPerc | tonumber > 80) | .Name'

Cleanup Procedures

Remove unused objects that retain memory-mapped caches. Target specific resource types rather than performing aggressive system-wide purges:

# Remove stopped containers older than 24 hours
docker container prune --filter "until=24h" --force

# Delete dangling and unreferenced images
docker image prune --all --filter "until=168h" --force

# Clean build cache and network overlays
docker builder prune --force

For comprehensive reclamation including volume data:

docker system prune --volumes --force

Service Recycling

When the Docker daemon itself exhibits resident set growth, restart the runtime without disrupting configured networks:

sudo systemctl reload-or-restart docker.service

On systems using SysVinit:

sudo service docker restart

Swap Configuration

Enable suppplementary swap space temporari to relieve memory pressure during high-load scenarios. First, allocate a 2 gigabyte swap file using a block-sized write:

sudo dd if=/dev/zero of=/swap-temp bs=1M count=2048
sudo chmod 600 /swap-temp
sudo mkswap /swap-temp
sudo swapon /swap-temp

To remove the swap file after the workload completes (which forces pages back into physical memory or clears them):

sudo swapoff /swap-temp
sudo rm -f /swap-temp

Verify current swap utilization with:

free -h | awk '/^Swap/ {print "Usage: " $3 "/" $2}'