Setup AWS Instance

Create security groups:

$ aws ec2 create-security-group \
   --group-name docker-machine \
   --description 'sc for running jupyter notebook server'
$ aws ec2 authorize-security-group-ingress \
   --group-name docker-machine \
   --protocol tcp \
   --port 22 \
   --cidr 0.0.0.0/0
$ aws ec2 authorize-security-group-ingress \
   --group-name docker-machine \
   --protocol tcp \
   --port 2376 \
   --cidr 0.0.0.0/0
$ aws ec2 authorize-security-group-ingress \
   --group-name docker-machine \
   --protocol tcp \
   --port 8888 \
   --cidr 0.0.0.0/0
$ aws ec2 authorize-security-group-ingress \
   --group-name docker-machine \
   --protocol tcp \
   --port 4567 \
   --cidr 0.0.0.0/0

Build the instance:

$ docker-machine -D create --driver amazonec2 \
	--amazonec2-instance-type p2.xlarge \
	--amazonec2-ami ami-2c57433b \
	--amazonec2-device-name "/dev/sda1" \
	--amazonec2-root-size "60" \
	--amazonec2-volume-type "gp2" \
	--amazonec2-region us-east-1 \
	--amazonec2-zone c \
	--amazonec2-retries 50 \
	--amazonec2-security-group docker-machine \
	--amazonec2-vpc-id $AWS_VPC_ID \
	--amazonec2-access-key $AWS_ACCESS_KEY_ID \
	--amazonec2-secret-key $AWS_SECRET_ACCESS_KEY \
	awsgpusd

The create usually fails and docker doesn't load correctly:

$ docker-machine stop awsgpusd
$ docker-machine start awsgpusd
$ docker-machine regenerate-certs awsgpusd

Check on instance status:

$ docker-machine ip awsgpusd
$ docker-machine inspect awsgpusd

Access the instance:

$ docker-machine ssh awsgpusd

Install official NVIDIA driver package:

# get repo setup
$ sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/7fa2af80.pub
$ sudo sh -c 'echo "deb http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64 /" > /etc/apt/sources.list.d/cuda.list'
$ sudo apt-get update && sudo apt-get install -y --no-install-recommends cuda-drivers

# force start the daemon
$ sudo nohup docker daemon -H tcp://0.0.0.0:2375 -H unix:///var/run/docker.sock &

Install nvidia-docker and nvidia-docker-plugin:

$ wget -P /tmp https://github.com/NVIDIA/nvidia-docker/releases/download/v1.0.1/nvidia-docker_1.0.1-1_amd64.deb
$ sudo dpkg -i /tmp/nvidia-docker*.deb && rm /tmp/nvidia-docker*.deb
$ sudo nvidia-docker run --rm nvidia/cuda nvidia-smi

Setup the environment:

$ eval `docker-machine env awsgpusd`
$ setenv NV_HOST "ssh://ubuntu@`docker-machine ip awsgpusd`:"
$ ssh-add ~/.docker/machine/machines/awsgpusd/id_rsa

Load local data into instance:

$ docker-machine scp -r . awsgpusd:

Create Docker Image

Build the image:

$ docker build -t kcavagnolo/docker_udsd:latest -f Dockerfile .
$ docker login -u $DOCKER_UNAME -p $DOCKER_PSWD
$ docker push $DOCKER_UNAME/docker_udsd:latest

Setup Jupyter notebook where env vars like $DOCKER_JUPYTER_PSWD are set to a value such as 'abc123':

$ nvidia-docker run \
  -it --rm \
  -p 8888:8888 \
  -v /home/ubuntu:/notebooks \
  -w /notebooks \
  -e PASSWORD=$DOCKER_JUPYTER_PSWD \
  $DOCKER_UNAME/docker_udsd

Clean-up

Shut-down the containers:

$ docker stop $(docker ps -a -q)
$ docker rm $(docker ps -a -q)

Close and kill instance:

$ docker rmi $(docker images -a -q)
$ docker-machine stop awsgpusd
$ docker-machine rm -f awsgpusd

nvidia-docker on AWS

Problem is launching from osx with nvidia-docker which cannot run on osx:

• https://github.com/NVIDIA/nvidia-docker/wiki/Deploy-on-Amazon-EC2
• NVIDIA/nvidia-docker#171

This works from osx command line:

$ docker run \
	--device=/dev/nvidiactl \
	--device=/dev/nvidia-uvm \
	--device=/dev/nvidia0 \
	-it \
	--rm \
	-p 8888:8888 \
	-v /home/ubuntu:/notebooks \
	-w /notebooks \
	-e PASSWORD=$DOCKER_JUPYTER_PSWD \
	kcavagnolo/docker_udsd

Maybe all-in-one solution?

$ sudo xhost +
$ sudo nvidia-docker run \
	--env="DISPLAY" \
	--volume="$HOME/.Xauthority:/root/.Xauthority:rw" \
	-env="QT_X11_NO_MITSHM=1" \
	-v /dev/video0:/dev/video0 \
	-v /tmp/.X11-unix:/tmp/.X11-unix:ro \
	-it \
	-p 8888:8888 \
	-p 6006:6006 \
	-v ~/sharefolder:/sharefolder \
	gtarobotics/udacity-sdc bash