Skip to content

Latest commit

 

History

History
executable file
·
295 lines (237 loc) · 13.9 KB

README.md

File metadata and controls

executable file
·
295 lines (237 loc) · 13.9 KB
Raw

rat-container

https://img.shields.io/badge/hosted-dockerhub-blue

Singularity and Docker recipes to build a SNO+ environment for RAT.

For regular usage, simply download the pre-built container with the following instructions for your container platform of choice. For advanced users, see the build instructions below.

IF THE DOCKERHUB LINK STOPS WORKING, SOMEONE MAY HAVE TO BUILD AND REUPLOAD THE CONTAINER TO DOCKERHUB DUE TO A CHANGE IN THEIR POLICY

As of November 1, 2020 Docker is implementing an inactive image removal policy, meaning in a free account (which is where this container is hosted) if the container is not updated or pulled for 6 consecutive months it will be deleted. This isn't a huge issue, someone will just have to do the following:

  • Build the container manually from the image file in this repository according to the instructions below
  • Upload it to another Dockerhub repository
  • Update the download links that reference the Dockerhub location with the new location

FEATURES

  • Full RAT-compatible environment, including ROOT5 (ROOT6 version now available), GEANT4 and scons
  • Can build any version of RAT
  • GUI output support on all operating systems
  • TensorFlow and CppFlow (CPU-only for the time being)
  • Singularity and Docker compatibility
  • *Cluster-compatible

*The image can be uploaded manually, pulled directly (if the cluster firewall permits) or run from /cvmfs; however, the cvmfs image is not always up-to-date with the repo version. This has been identified as an issue with a possible solution posed.

[PLEASE READ]

  1. Singularity and Docker are similar tools but operate slightly differently. Singularity acts more like an overlay, where you have access to your filesystem as you would outside the container (with the same rights as you'd have outside), whereas Docker provides you with an isolated virtual filesystem (meaning you can't access your files from outside the container). In summary, it is best to mount whatever directories you may need when running the container, whether in Docker or Singularity (see the section "To write/execute files from directories outside of RAT/launch directory" below).

  2. Regardless of whether you download or build the container, you can use and develop RAT as you see fit as it is external to the container.

  3. Instructions to install Singularity can be found here. For Docker, instructions for each platform can be found here.

  • For Singularity, version 3.2+ is required
  • For Docker, version 19.0+ is required

  1. As the DIRAC system no longer supports SL6, there is no longer a need to maintain an SL6 version when pushing new RAT releases to cvmfs. Therefore, the only image offered here is based on SL7.

  2. To be clear, if you wish to use the prebuilt image, then you do NOT need to clone this repo; simply follow the instructions below.

New Video Tutorial (slightly outdated - no longer necessary to source the setup-env.sh on startup)

To download the pre-built container

If on a shared system/cluster, Singularity should be available so use the following command to obtain the latest version of the container:

singularity pull --name rat-container.sif docker://snoplus/rat-container:root5

The tag (in the above command, root5) can be replaced with the desired tag.

Ensure that the Singularity version you are using is ≥3.2

At the moment, certain clusters (like Cedar) have firewall rules preventing access to SingularityHub. There is a version of the image located at /cvmfs/snoplus.egi.eu/sl7/sw/containers/rat-container.sif but keep in mind that it may not always be the latest version (this shouldn't matter if you are simply building/running RAT).

If on your own local machine, Docker should be used as it is easier to install. The command to obtain the latest version of the container for Docker is:

docker pull snoplus/rat-container:root5

The tag (in the above command, root5) can be replaced with the desired tag.

Docker doesn't actually create a file in your working directory in the same way that Singularity does; rather, it downloads the image layers and adds an entry to your local Docker registry which can be viewed by going:

docker images

This difference doesn't have an effect on how the container is actually used.

Instructions on how to use the container with RAT

To build RAT for the first time:

  • Clone RAT from GitHub (NOTE - If on Windows, make sure you run git config --global core.autocrlf input prior to cloning or else Git will automatically change the Unix line-endings to Windows (which will break the next steps)

  • Enter the following command, filling in the path to RAT with your own. This will mount your RAT repo to the directory /rat inside the container:

    For Singularity:

    singularity shell -B path/to/rat:/rat rat-container.sif

    For Docker:

    docker run -ti --init --rm -v /absolute/path/to/rat:/rat snoplus/rat-container

    Note - the -v flag operates the same as -B in Singularity BUT you must provide it with an absolute path (one starting at /); relative paths (the path from where you are now) will not work.

  • Once in the container, Singularity users need to run the following:

    source /home/scripts/setup-env.sh

    In Docker this is unnecessary as Docker sources it automatically on launch. You may see a message about how it could not find /rat/env.sh; this is expected as you have not built RAT yet. If the build is successful, you shouldn't see this message next time.

    • Finally, run this command to build RAT:
    source /home/scripts/build-rat.sh

    Alternatively, scons can manually be called while in the /rat folder.

  • RAT is now ready to use! Look at the instructions below for how to run it

To exit the container (Singularity and Docker):

exit

To run RAT:

  • First, get a shell into the container with your RAT bound into it: (It is important to mount your rat directory to /rat as the build scripts look there for it!)

    For Singularity:

    singularity shell -B path/to/rat:/rat rat-container.sif

    For Docker:

    docker run -ti --init --rm -v /absolute/path/to/rat:/rat snoplus/rat-container

  • RAT is now ready for use, and you should be able to access the RAT repo itself at /rat. To use other directories, additional bind mounts are necessary (see below).

To use GUI apps like ROOT's TBrowser: (This is based on CERN's documentation for running ROOT with graphics)

  • The process is different on each OS but I will outline steps here to make it work on each. Note that these instructions assume that since you are on your own machine, you are using Docker. Singularity may work with graphics as it is, but these Docker solutions are the only ones that are tested and confirmed to be working.

    For Linux:

    docker run -ti --init --rm --user $(id -u) -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix -v /absolute/path/to/rat:/rat snoplus/rat-container

    As you can see, the difference is a few extra options. As the command has gotten so large, you can set an alias in your .bashrc to something much shorter and more convenient.

    For Windows 10:

    As of the new May 2020 Windows update, the Windows Subsystem for Linux (WSL) version 2 is out. Docker desktop can be configured to use this which is the recommended way to run Docker on Windows. Ensure WSL2 is enabled in the Docker Desktop settings, then follow these instructions:

    1. Download and install Xming
    2. When Windows prompts you to allow it in the firewall, do so.
    3. Finally, restart Xming and now run the following command in Powershell or WSL2:
    docker run --init --rm -ti -e DISPLAY=host.docker.internal:0 -v /absolute/path/to/rat:/rat snoplus/rat-container

    For macOS:

    1. Install XQuartz
    2. Open XQuartz, and then go XQuartz -> Preferences -> Security, and tick the box "Allow connections from network clients"
    3. Run xhost + 127.0.0.1 which will whitelist your local IP
    4. Finally, you can run the container with the following:
    docker run --rm --init -ti -v /tmp/.X11-unix:/tmp/.X11-unix -v /absolute/path/to/rat:/rat -e DISPLAY=host.docker.internal:0 snoplus/rat-container

    (The order -ti instead of -it seems to only matter for MacOS)

To update RAT:

  • Outside of the container, cd into your RAT repo, and run:

    git pull origin master

  • Then, run the container:

    For Singularity:

    singularity shell -B path/to/rat:/rat rat-container.sif

    For Docker:

    docker run -ti --init -v "$(pwd)"/rat:/rat snoplus/rat-container

  • Navigate to the RAT directory:

    cd /rat

  • Finally, run the build script (/home/scripts/build-rat.sh) or scons directly to rebuild RAT:

    scons

To write/execute files from directories outside of RAT/launch directory:

  • Add additional bind mounts to your Singularity or Docker command

  • Example:

    For Singularity:

    singularity shell -B path/to/rat:/rat,/other/path:/stuff rat-container.sif

    For Docker:

    docker run --init --rm -ti -v /absolute/path/to/rat:/rat -v /other/path:/stuff snoplus/rat-container

  • Now in the container, you have access to /other/path at /stuff

To use a specific branch of RAT:

  • Ensure you git checkout to the branch OUTSIDE the container to avoid issues, then run RAT like above

To use TensorFlow/cppflow:

  • The libraries are already installed (tensorflow at /usr/local/lib, cppflow repo is at /home/software) and the environment variables are set in the setup-env.sh script, so you should be able to just use it after sourcing

[ADVANCED]

To build the container

To build, you must have root permissions and Docker installed on your machine. Docker installation instructions can be found here for each OS.

To rebuild the container:

  1. Clone this repository

  2. Navigate into either /ROOT5 or /ROOT6 depending on which you would like to build off of

  3. Edit Dockerfile, which is the recipe on what you would like to put into your container

  4. Once you are happy with your changes, navigate back to the root of the repository and run:

    docker build -t YOUR_CONTAINER_TAG -f ROOT5/Dockerfile .

    where YOUR_CONTAINER_TAG is the name you would like to give to your container. Also, ensure you change ROOT5 to ROOT6 if using that version

  5. This will build your container with your tag name, which you can then use in the same way as in the above guide, but instead of

    docker run ... snoplus/rat-container

    you will now run:

    docker run ... YOUR_TAG_NAME

  6. [OPTIONAL] If you would like to share or back up your container image, you can push it to Dockerhub. You can follow the official documentation to learn how

To run multiple RAT instances

If you want to use multiple RAT instances simultaneously, then all you have to do is run an instance of this container with each version of RAT that you want; do NOT try mounting multiple RATs to the SAME instance as the image was not configured for this.

To modify Geant4

If you need to edit Geant4 for any reason, you will have to modify the recipe file and make your changes accordingly, then rebuild the container.

F.A.Q.

On macOS I see "docker: Error response from daemon: Mounts denied: The path ... is not shared from OS X and is not known to Docker."

  • This happens because Docker only allows mounting from 4 locations by default to follow Apple's sandbox guidelines; these locations are: 
    /Users
/tmp
/private
/Volumes
  • Ensure your RAT repository is stored in one of these locations (the easiest would be simply under /Users/[your username]/rat)

I'm seeing "/usr/bin/bash: /usr/bin/bash: cannot execute binary file" when I try to run the container

  • This happens because you have bash at the end of your run command; in the new version, this is no longer necessary as it will launch bash by itself.

I'm seeing "Error getting image manifest using url..." when I try to pull the container

  • This seems to happen on the clusters, most likely due to the firewall. Try pulling the container on your local machine, and transfer the image to your cluster with scp.

I'm seeing errors when running scons to rebuild RAT after updating to a new RAT release

  • This happens when you use the GUI-enabled docker command (not the standard command) when launching the container to rebuild RAT. Please review the instructions for how to update RAT above for the correct way to update.
  • This can also happen if you don't run scons within the /rat directory as it won't be able to find the correct files

When I try to open the TBrowser/another GUI app, it doesn't show

  • This is a known issue, and happens for two reasons. If you are trying to use the Docker version on your own machine, Docker does not have access to the display by default so there is some configuration required.
  • The other issue is if you are trying to do this on a cluster with the Singularity version, you will notice the same thing. Because you are remotely connected, the display is not configured by default to also connect.
  • Known methods for getting a GUI working are listed in a section above for each OS under Docker.