Intel Xeon服务器Docker资源扩容管理及Cloudera容器Solr端口问题咨询

Hey there! Let's work through your three main issues one by one—you've got a powerful server, so we can definitely optimize it to run more containers, fix the resource saturation problem, and get that Solr port working.

Your server has plenty of CPU (32 cores) and memory (~1TB), so the bottleneck is likely Docker's default resource constraints or per-container resource allocations. Here's how to adjust:

• Tweak the Docker Daemon's Global Resource Limits
  Edit (or create) the /etc/docker/daemon.json file to set global constraints and optimize resource usage:

  {
    "default-cpu-shares": 1024,
    "default-memory": "32g",
    "default-memory-swap": "64g",
    "storage-driver": "overlay2",
    "max-concurrent-downloads": 10,
    "max-concurrent-uploads": 10,
    "live-restore": true
  }
  

  • default-memory: Sets a default memory cap per container (adjust based on what each Cloudera quickstart actually needs—if each only uses 16GB, you can lower this to fit more containers).
  • default-cpu-shares: Controls CPU priority (1024 is the baseline; lower values mean containers get less CPU time when under load).
    After editing, restart Docker with systemctl restart docker.

• Adjust Kernel Parameters for Higher Container Density
  Increase system limits to handle more containers:

  • Edit /etc/sysctl.conf and add these lines:

    fs.file-max = 1000000
    net.ipv4.ip_local_port_range = 1024 65535
    net.core.somaxconn = 4096
    vm.max_map_count = 262144
    

  Apply changes with sysctl -p. These settings increase open file limits, available ports, and memory map counts—critical for running dozens of containers with multiple services.

• Set Per-Container Resource Limits Explicitly
  When creating new Cloudera containers, specify exact CPU/memory limits instead of letting them take unlimited resources. For example:

  docker run -d --name cloudera-quickstart-XX \
    --cpus="2" \
    --memory="16g" \
    --memory-swap="32g" \
    -p 8888:8888 -p 8983:8983 \
    cloudera/quickstart:latest
  

  Calculate based on your total resources: 32 cores / 2 cores per container = 16 containers on CPU alone, but with CPU sharing (since not all containers will max out CPU at once), you can safely run more. For memory: 1TB /16GB per container = 62 containers, which hits your 50+ goal.

If Docker is already using all your server's resources, here's how to free up space and optimize:

• Clean Up Unused Docker Resources
  Run these commands to prune dead containers, unused images, and dangling volumes:

  docker container prune -f
  docker image prune -a -f
  docker volume prune -f
  

  This will remove the 7 stopped containers and any unused assets eating up disk or memory.

• Pause or Stop Non-Critical Containers
  If you need immediate space to create new containers, pause running ones that aren't in use:

  docker pause <container-id>
  

  Or stop them temporarily (you can restart later):

  docker stop <container-id>
  

• Enable Docker Swarm for Better Resource Scheduling
  For larger container fleets, set up Docker Swarm to distribute resources across your server more efficiently. It will automatically schedule containers based on available CPU/memory, preventing any single container from hogging resources.

• Check for Resource Leaks
  Use docker stats to identify containers using more resources than expected. If any have memory that keeps growing indefinitely, they might have leaks—restart them or check the Cloudera service logs inside the container for issues.

Since Hue works but Solr doesn't, let's narrow down the problem:

• Verify Solr is Running Inside the Container
  Exec into the container and check the Solr service status:

  docker exec -it <container-id> service solr status
  

  If it's stopped, start it with service solr start and check logs for errors:

  docker exec -it <container-id> tail -f /var/log/solr/solr.log
  

  Common issues here: missing dependencies, port conflicts inside the container, or incorrect Solr configuration.

• Check Port Mapping and Host Firewall
  Run docker ps to confirm the Solr port (default 8983) is mapped correctly to the host. If it's missing, re-run the container with -p 8983:8983 (or a different host port if 8983 is in use).
  Also, ensure your server's firewall allows incoming traffic on the mapped port:

  iptables -A INPUT -p tcp --dport 8983 -j ACCEPT
  

• Ensure Solr is Bound to All Interfaces
  Inside the container, check Solr's jetty configuration to make sure it's not only listening on localhost. Edit /opt/cloudera/parcels/CDH/lib/solr/server/etc/jetty.xml and look for the <Set name="host"> line—set it to 0.0.0.0 instead of 127.0.0.1, then restart Solr.

• Check for Host Port Conflicts
  Verify the host port you're using for Solr isn't taken by another service:

  netstat -tulpn | grep 8983
  

  If it's in use, map Solr to a different host port (e.g., -p 8984:8983) when starting the container.

内容的提问来源于stack exchange，提问作者issa hamieh