Docker tutorial

The official installation guide is available here. For Windows and Mac, binaries are available to be downloaded. For Linux, you need to choose your distribution and follow the guide for installation from repository (recommended), .deb package, or with a convenience script.

• Download the image. The tag can be one of full, icub, graspit, hri. Tag full contains iCub, grasping and HRI related things. Tag icub contains only iCub related software, tag graspit contains only grasping related things and tag hri contains HRI related software.

   docker pull b3m33hro/labs:tag

• Name your image for an easier usage. <image_name> is an arbitrary name for your image, e.g., hro_image

   docker tag b3m33hro/labs:tag <image_name>

• Run the image. -it is for interactive mode. -p determines which ports will be accessible from the host system. <container_name> is an arbitrary name of your container, e.g., hro. -v specifies which folder will be shared between the container and the host system (optional). <image_name> is the name selected in the previous step.
  • The shared folder is used for easier transfer of files created in the Docker container. You can also share code that way, so you can edit it from both the host machine and from inside the container.

   docker run -it -p 6080:6080 -p 8888:8888 --name <container_name> -v path_to_data_folder_host_system:/root/data <image_name>

• From inside the container run command to start the VNC server. The server is needed for OpenGL software.

   start-vnc-session.sh

• On your host machine open any internet browser and type http://localhost:6080 to the address bar. A window with a Linux environment should open.
  • The VNC server may crash sometimes. To fix it, just run start-vnc-session.sh again.
  • Unfortunately, you can not copy-paste from the host machine to the VNC.
    • Solution is to copy things from the internet browser directly in the VNC. Another possibility is to use the VNC only as a “screen” and use a terminal on the host machine to control it.

• Choregraphe can be run from terminal with the command choregraphe
  • you can select 90-days test, or enter this key 654e-4564-153c-6518-2f44-7562-206e-4c60-5f47-5f45

• RoboDK must be installed first (because of licensing). One can do it from terminal with command /root/Install-RoboDK. The program can be then run with command /installation_path/RoboDK-Start.sh (where installation_path is /RoboDK in default), or with Desktop icon.
  • You will have 30days licence (which should be enough)

• Stop the container (any of the three methods will work)

   docker stop <container_name>

   CTRL+D

   exit

• Restart the container

   docker start <container_name>  
   docker attach <container_name>

• Open another terminal in the same container

   docker exec -it <container_name> /bin/bash

• Remove a stopped container from your system

   docker container rm <container_name>

• Clean up all docker resources

   docker system prune -a

• Check GraspIt!. Open the terminal and run graspit_simulator
• Check Jupyter. Open terminal, run jupyter-notebook. Open a browser (in container or in the host machine) and enter http://localhost:8888 and you should see Jupyter notebook interface.
  • Do not run the notebook from Applications-Development menu

• Check Robotology build. Open the terminal and run the following lines. Gazebo interface with iCub robot should appear.

   git clone https://gitlab.fel.cvut.cz/body-schema/teaching/b3m33hro-labs.git
   cd b3m33hro-labs/lab3/tutorial_cartesian-interface/smoke-test
   ./test.sh

In case of problems with Docker on your computer, you can use Docker on Gitpod. Gitpod allows running Docker images on their servers for free (each registered user can use 50 hours per month).

• Open Gitpod
  • you will need github.com account
  • Image should be automatically downloaded and run in your new Gitpod workspace
  • This will open the tag full, check other branches of https://github.com/rustlluk/B3M33HRO- * gitpod to open other tags

• No docker run and others are needed
• Try Functionality check again
• With every new terminal you should be automatically switched to user root. If not, please run su with password root.
  • In older version of the image, you need to run su as the first command in each new terminal you open and use the environment as before
    • password is root

Everything should work almost the same as in the local installation, but server resources are used. However, you can not link folder from your PC as in the case of local Docker. You can either use data only in docker, or there are several ways how to share code:

• Always upload and download data from the main page of Gitpod workspace
  • the workspace is in location /workspace and linked to /root/workspace
• Use Git. You can:
  • Run Gitpod from your repository by:
    • make a fork of https://github.com/rustlluk/B3M33HRO-gitpod, or
    • create a new repository and add .gitpod.yml to it.
  • Use other repository to share code, e.g., fork of our lab repository.
• Mount some folder on public server (e.g., your folder at taylor.felk.cvut.cz) to the Docker environment and your host PC (with sshfs or other). This option may work almost the same as -v option in local Docker installation.

You can copy/paste things to the Gitpod workspace terminals, but you need to approve it (popup window with the question should appear after you first try to paste something).

• On Windows, the Docker sometimes allocates too much memory, which may cause Gazebo to fail. Restart the Docker (from status bar) or restart the computer.
• If you want to install anything inside the Docker, apt update will most probably be needed to be run first. Else you will get no package … error.
• If you are using Mint, do not use $(lsb_release -cs) when adding the repository, but use name of Ubuntu version corresponding to your Ubuntu kernel from Ubuntu versions.
• You can also run the smoke-test without GUI, which may be faster on slower computers. Just open some_path/smoke-test/fixtures/fixture.xml in the given project you want to test and remove/comment the whole <module> section containing gzclient.
• Try to run Gazebo at least twice, it may fail sometimes.
• If your Gazebo fails in the middle of operation try to free as many resources on your computer as possible (RAM, CPU) by, for example, closing other internet browser tabs. Restart Docker, or try to kill other running processes of yarp or Gazebo.
  • check whether anything is running in top or htop and kill it (look for anything with yarp, gz, iCartesian* etc.). However, using pkill -f gz, pkill -f yarp and pkill -f tutorial should be enough in most cases to kill all Gazebo and Yarp related things.

You need to install Robotology superbuild, GraspIt!, gpd, Python 2.7 and 3.X. Optionally, ROS is needed. You can check Dockerfile to get an idea of how everything is installed.