ltl7155/create_cuda_containers

Install docker and nvidia-container-toolkit

Install nvidia-container-toolkit according to the installation guide

sudo apt-get install nvidia-docker2

sudo usermod -aG docker kangjie

restart your terminal

Build image

docker pull nvidia/cuda:11.2.1-cudnn8-devel-ubuntu20.04 or other images

vim Dockerfile

docker build -t kangjie/contianer_image:cuda11.2.1-cudnn8-devel-ubuntu20.04 .

Download Miniconda

cd init_files/

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

Create a new container

sudo ./create_containers.sh

Add a user for the container

inside the container

docker exec -it container_name /bin/bash

cd /home/user_name/

sudo bash add_user.sh

exit

outside the container, test for the create user with the specific port

ssh user_name@ip -p port

bash Miniconda.sh

exit

login without password

inside the local machine

scp -P 32869 ~/.ssh/id_rsa.pub [email protected]:~/

# ~/.ssh/config
Host  server6kangjie
    HostName   155.69.148.30
    Port       32869
    User       kangjie
    IdentityFile    ~/.ssh/id_rsa

inside the container

mkdir ~/.ssh/

cat id_rsa.pub >> .ssh/authorized_keys

Check config for a container

check container first and remeber the container ID

docker ps

stop docker service

sudo systemctl stop docker

check config file for the container

sudo ls /var/lib/docker/containers/

sudo vim /var/lib/docker/containers/container-id/config.v2.json

start docker service

sudo systemctl start docker