Docker

Usually the first issue is running docker without admin rights

Please refer to this

If you want to attach your shell to a running container try this

$ docker ps # use that to get the name of the running container (i.e. "cranky_dirac")
$ docker exec -it <container name> /bin/bash  # attach shell

Commit changes to an image

When we modify an image (i.e install dependencies) and want the image to retain it’s modified state.

• Identify image: $ docker ps (hint: zoom the image below :p)

• Commit Changes docker commit -m "your-commit-message" eager_hodgkin nvcr.io/nvidia/tensorflow:21.10-tf1-py3-ganindu_matplotlib

Docker on Nvidia Jetson

Setting up the runtime

Important prerequisite: sudo vim /etc/docker/daemon.json

{
    "default-runtime": "nvidia",
    "runtimes": {
        "nvidia": {
            "path": "nvidia-container-runtime",
            "runtimeArgs": []
        }
    }
}

1. Connect all devices and cables.
2. Connect power to the Device
3. Connect power to additional equipment that needs separate power (if applicable).

4. set the “DISPLAY” env variable if you are using a remote shell (e.g. export DISPLAY=:1) if this does not work run xhost + in the jetson to allow remote displays. (xhost + can only run directly via a keyboard attached to jetson for security reasons)

5. test display functionality with xeyes: (There should be an output on the screen attached to the jetson
6. in a shell in jetson cd /Workspace/your/docker/folder/with/the/compose/file
7. docker-compose up
8. sudo xhost +si:localuser:root # if you use this you’ll have to export the display variable e.g. export DISPLAY=:1
9. docker exec -it <the-container-name> /bin/bash --login
10. try if xeyes work (optional)
11. do your stuff
12. After running do xhost - to shut down access to xhost by anyone

TIPS FOR ACESSING DOCKER CONTAINER FROM A REMOTE:

Remember to do docker-post installation steps

Create the docker group (If already not done).

• run sudo groupadd docker

Add your user to the docker group.

• run sudo usermod -aG docker $USER

Docker and Jupyter notebooks

if you haven’t installed already use

python3 -m pip install ipykernel to install the ipykernel python3 -m pip install jupyter to install Jupyter

because you used python install the PATH variable may not be updated so try using python3 -m jupyter notebook .

Nvidia Userguide https://docs.nvidia.com/deeplearning/frameworks/user-guide/index.html

Migrating docker storage

Sometimes to use the available disk space optimally we need to move things around and docker storage directory is no exception, as of writing docker by default uses the /var/lib/docker directory. let’s say we mounted our additional storage onto /mnt/my/new/storage and we want to place the docker storage in the /docker/storage directory relative to that.

orignal_source

STEP 1

stop docker (you can pick when perform this action) sudo service docker stop

STEP 2

create the ‘daemon.json’ on /etc/docker (verify with docker docks that this is the correct location)

 sudo vim /etc/docker/daemon.json

and then edit according to your use case.

{
    "data-root": "/mnt/my/new/storage/docker/storage"
}

STEP 3

Copy the files to the destination

sudo rsync -vaP /var/lib/docker /mnt/my/new/storage/docker/storage

STEP 4

we can rename the old directory to a make sure the changes have taken effect.

sudo mv  /var/lib/docker /var/lib/docker.old

STEP 5.

Start the docker service again to see if everything works

sudo service docker start

IMPORTANT

• Doing this with debilitating hardware constraints can lead to run time difficulties in execution.
• Be aware the that an active docker container uses more space in /var/lib/docker (you can confirm this with a command such as du -h d 1 before and after activating a container.).