Introduction

We would like to explore the potential of docker in our community. This repository is there a first attempt in this direction.

Main use case

I have a Mac laptop and I would like to work locally on my Mac in a UNIX system for the following reasons:

• to work directly on lxplus can be slow (node dependent). The advantage of lxplus is that it mounts afs, eos, cvmfs... and the HTCondor interface.
• work on SWAN can be slow and it does not mount afs - this can be problematic for our MAD-X mask.
• python C-compiled package can have difficulties to be maintained for Mac/Win, e.g., cpymad does not work properly on Mac (I manage to install it on Mad and run small lattice, but when I try the larger ones, namely LHC, it fails). I started to use a simple docker as a workaround (the same I used for CAS/JUAS courses), typically I use jupyter lab and my experience in term of performance and stability is positive.
• I would like to mount eos and afs without suffering to much
• I would like that the possibility to share my setup not only via a set of wikis but also with a press-one-button solution. With this spirit, one could prepare and share several conda env, e.g,. to track particles, to access NXCALS,...
• one can also couple VSCode directly on the container.

For the moment, I am considering that you have already installed in your Mac (localhost) afs - it is quite easy to do so, I tested the docker only on Catalina. You do not need to install eos on your Mac. This is installed inside the container (thanks to the help of Fabio Luchetti).

To run the docker

To lauch the docker sterbini/test

docker run -it -u 0 -v /afs:/afs -v ~:/home/jovyan/local_host_home -p 8888:8888 --rm --cap-add SYS_ADMIN --device /dev/fuse gitlab-registry.cern.ch/abpcomputing/sandbox/be-abp-docker:$commit_tag

where you can find the available $commit_tag at https://gitlab.cern.ch/abpcomputing/sandbox/be-abp-docker/container_registry/9443. Probably you want the latest tag.

At the time of the writing (2021.05.07) the latest tag was 288bbf2a.

In case you want to mount a local folder, this has to be shared with Docker Docker (I did it with Docker Desktop on my Mac) and then use the -v option

docker run -it -u 0 -v /afs:/afs -v ~:/home/jovyan/local_host_home -p 8888:8888 --rm --cap-add SYS_ADMIN --device /dev/fuse gitlab-registry.cern.ch/abpcomputing/sandbox/be-abp-docker:288bbf2a

The volumes you mount are very important, since only the information you save there will survive after you exit the container.

When the container is starting will try to run the file /home/jovyan/local_host_home/docker_start.sh. You can customize it accordingly to your need. You find a template of the file in /home/jovyan/docker_start.sh. So please copy it on /home/jovyan/local_host_home.

You can run a jupyter notebook server with jupy.

When you are in the docker you have to do:

kinit $YOUR_NICE_LOGIN_NAME

To make the docker

To build the docker

docker build . --build-arg username=jovyan -t $IMAGE_NAME

where $IMAGE_NAME can be for example sterbini/test

To export the X-graphics on Mac

From https://medium.com/@mreichelt/how-to-show-x11-windows-within-docker-on-mac-50759f4b65cb

• Install the latest XQuartz X11 server and run it
• Activate the option ‘Allow connections from network clients’ in XQuartz settings
• Quit & restart XQuartz (to activate the setting)

On the local host (Mac) do

xhost + 127.0.0.1

And when you run the docker by adding

-e DISPLAY=host.docker.internal:0

to the docker run that is

docker run --rm -e DISPLAY=host.docker.internal:0 -p 8889:8888 -ti --cap-add SYS_ADMIN --device /dev/fuse -e GRANT_SUDO=yes gitlab-registry.cern.ch/abpcomputing/sandbox/be-abp-docker:$TAG

then you can test from the container

xclock

Some simple operations with Docker

To see all running containers you can do (https://phoenixnap.com/kb/how-to-ssh-into-docker-container)

docker ps

then you can attach from another local-host terminal to the container

docker attach $CONTAINER_ID

the $CONTAINER_ID is the list of the docker processes - you can also use the $CONTAINER_NAME to attach it.

If you use docker attach you are attaching the very same terminal. If you want to open a new terminal in the container you can do

docker exec -it $CONTAINER_ID /bin/bash

Containers renaming, stopping and reattaching.

You can use '-n' option to name a container (no other containers with that name should existin in the Docker Host). Then

docker run --name happy_newton -e DISPLAY=host.docker.internal:0 -p 8889:8888 -p 8822:22 -ti --cap-add SYS_ADMIN --device /dev/fuse -e GRANT_SUDO=yes -v /afs:/afs $MY_IMAGE:$MY_TAG

One you exit the container, it will not be removed but it will stopped. You can remove it automatically if you add --rm option

docker run --rm --name happy_newton -e DISPLAY=host.docker.internal:0 -p 8889:8888 -p 8822:22 -ti --cap-add SYS_ADMIN --device /dev/fuse -e GRANT_SUDO=yes -v /afs:/afs $MY_IMAGE:$MY_TAG

It you do not remove the container when you exit it you can restart it and reattach it. For example:

docker container start happy_newton
docker attach happy_newton

You can also remove a

container rm happy_newton