Allocating more memory to Docker Enterprise Preview Edition

Question

I am running Docker Enterprise Preview Edition on Windows Server 2019 and have managed to pull and run the docker-compose.yml file below. However shortly afterwards the container shuts down and when I run the command docker-compose logs it shows me the insufficient memory issue below:

Docker-compose file

version: '3.7'

services:

  elasticsearch:
   container_name: elasticsearch
  #  image: docker.elastic.co/elasticsearch/elasticsearch:7.9.2
   image: docker.elastic.co/elasticsearch/elasticsearch:7.17.1
   deploy:
       resources:
          limits:
            cpus: 0.25
            memory: 4096m
   ports:
    - 9200:9200
   volumes:
    - C:\DockerContainers\Elasticsearch\data:/usr/share/elasticsearch/data
    - C:\DockerContainers\Elasticsearch\config\certs:/usr/share/elasticsearch/config/certs
   environment:
    - xpack.monitoring.enabled=true
    - xpack.watcher.enabled=true
    - ES_JAVA_OPTS=-Xms512m -Xmx512m
    - discovery.type=single-node
  #  networks:
  #   - elastic



  kibana:
   container_name: kibana
  #  image: docker.elastic.co/kibana/kibana:7.9.2
   image: docker.elastic.co/kibana/kibana:7.17.1
   deploy:
       resources:
          limits:
            cpus: 0.25
            memory: 4096m   
   ports:
    - 5601:5601
   volumes:
    - C:\DockerContainers\Elasticsearch\Kibana\config\certs:/usr/share/kibana/config/certs    
   depends_on:
    - elasticsearch
  #  networks:
  #   - elastic

# networks:
#   elastic:
#     driver: nat

Docker logs

elasticsearch    | # There is insufficient memory for the Java Runtime Environment to continue.
elasticsearch    | # Native memory allocation (mmap) failed to map 65536 bytes for committing reserved memory.
elasticsearch    | # An error report file with more information is saved as:
elasticsearch    | # logs/hs_err_pid7.log

I read on the elasticsearch Docker guideline that it needs at least 4GB RAM. I have included the RAM limit in the docker compose yml file but it doesn't seem to take effect. Does anyone know how to set the memory usage for Docker which is running on Windows Server 2019?

Answer 1

I ran into this same issue trying to start Docker using almost exactly the same configuration as you, but doing so from Windows 10 instead of Windows Server 2019. I suspect the issue isn't the memory configuration for the Elastic containers, but rather the Docker host itself needs to be tweaked.

I'm not sure how to go about increasing memory for the Docker host when running Docker Enterprise Preview Edition, but for something like Docker Desktop, this can be done by changing the memory limit afforded to it by adjusting the Settings -> Advanced -> Memory slider. Docker Desktop may need to be restarted afterwards for this change to take effect.

Similarly, if you are running a Docker machine, like I am, it may be necessary to recreate it, but with more than the default 1GB that is allotted to new Docker machine instances. Something like this:

docker-machine create -d vmwareworkstation --vmwareworkstation-memory-size 8192 default

Swap in whatever vm type makes sense for you (e.g., VirtualBox) and remember that you need to run docker-machine env | Invoke-Expression in each console window in which you intend to run Docker commands.

I saw definite improvement after giving the Docker host more breathing room, as the memory-related errors disappeared. However, the containers still failed to start due to the following error:

max virtual memory areas vm.max_map_count [65530] is too low, increase to at least [262144]

This is a known issue (see https://www.elastic.co/guide/en/elasticsearch/reference/current/docker.html#_set_vm_max_map_count_to_at_least_262144). Again, if you're using Docker Desktop with WSL2 installed, you can run this:

wsl -d docker-desktop sysctl -w vm.max_map_count=262144
sysctl -w vm.max_map_count=262144

For a Docker machine, you'll need to ssh into it and set the vm.max_map_count directly:

docker-machine env | Invoke-Expression
docker-machine ssh default
sudo sysctl -w vm.max_map_count=262144

The above change to the Docker machine will last during your current session but will be lost after reboot. You can make it stick by:

1. adding sysctl -w vm.max_map_count=262144 to the /var/lib/boot2docker/bootlocal.sh file. Note this file may not exist before your changes.
2. running chmod +x /var/lib/boot2docker/bootlocal.sh
3. exit ssh and restart the Docker machine via docker-machine restart default.
4. to confirm the change, run docker-machine ssh default sudo sysctl vm.max_map_count. You should see it set to 262144.

After these changes, I was able to bring the elastic containers up. You can smoke test this by issuing a GET request for http://localhost:9200 in either Postman or curl. If you're using Docker machine, the ports you've set up are accessible to the machine, but not to your Windows box. To be able to connect to port 9200, you'll need to set up port forwarding via the docker-machine ssh -f -N -L 9200:localhost:9200 command.

Also, consider adding the following to your Docker compose file:

environment:
  - bootstrap.memory_lock=true
ulimits:
  memlock:
    soft: -1
    hard: -1

Hope this helps!