udocker - be anywhere

Part 2 - Hands On: basic stuff

https://github.com/indigo-dc/udocker

Mario David david@lip.pt, Jorge Gomes jorge@lip.pt

---

What udocker is not - I

• Not appropriate to run services:
  • In most cases you need root privileges to run services.
  • You have Docker (or other container tools) for this.
• udocker is a run-time and is not meant to build docker images:
  • Docker images should be built with Docker.
  • Use you (Lap/Des)top with Docker, for this.

---

What udocker is not - II

• docker-compose like functionality:
  • This is usually to compose micro-services to deploy a platform/service.
  • Again udocker is not appropriate to run services.
  • Use docker-compose itself for this.

---

udocker aims/objectives

• Execute applications encapsulated with dependencies in containers
  • as non privilege user.
• Execute containers from docker images
  • including officially supported images in Dockerhub.
• Execute applications with very specific, customized libraries and environments
  • difficult to support in very controlled systems such as HPC machines.

---

udocker: Installation

https://indigo-dc.github.io/udocker/installation_manual.html

---

Installation: tarball

Access the INCD advanced computing facility at Lisbon using ssh:

ssh -l <username> cirrus8.a.incd.pt
module load python/3.10.13

• The end user can download and execute udocker without system administrator intervention.
• Install from a released version:
  • Download a release tarball from https://github.com/indigo-dc/udocker/releases:

wget https://github.com/indigo-dc/udocker/releases/download/1.3.10/udocker-1.3.10.tar.gz
tar zxvf udocker-1.3.10.tar.gz
export PATH=`pwd`/udocker-1.3.10/udocker:$PATH

---

Installation: PyPI - I

• Install from PyPI using pip:
  • For installation with pip it is advisable to setup a Python3 virtual environment

python3 -m venv udockervenv
source udockervenv/bin/activate
pip install udocker==1.3.10

---

Installation: PyPI - II

The udocker command will be udockervenv/bin/udocker.

• Optionally, we can set UDOCKER_DIR environment variable where the binaries, libraries images and containers will be saved. The default directory is $HOME/.udocker.

mkdir udocker-tutorial
cd udocker-tutorial/
export UDOCKER_DIR=$HOME/udocker-tutorial/.udocker

(More details: https://indigo-dc.github.io/udocker/installation_manual.html)

---

Installation: tools and libraries - I

• udocker executes containers using external tools and libraries that are enhanced and packaged for use with udocker.

• To complete the installation, download and install the required tools and libraries.

udocker install

---

Installation: tools and libraries - II

• Installs by default in $HOME/.udocker, or in UDOCKER_DIR=$HOME/udocker-tutorial/.udocker.

• Explore the directory structure under $HOME/udocker-tutorial/.udocker

---

udocker: CLI - the basic (introductory) stuff

https://indigo-dc.github.io/udocker/user_manual.html

---

0. help and version

Global help and version

udocker --help
udocker --version

You can get help on a given command

udocker run --help

---

1. pull

Pull an image from Dockerhub (for example, an officially supported tensorflow):

udocker pull tensorflow/tensorflow

---

2. images

List the images in your local repository (-l option shows long format):

udocker images
udocker images -l

---

3. create

To create a container named mytensor, the default execution engine is P1 (PTRACE + SECCOMP filtering):

udocker create --name=mytensor tensorflow/tensorflow

---

4. ps

List extracted containers. These are not processes but containers extracted and available for execution:

udocker ps

---

5. run: I

Executes a container. Several execution engines are provided. The container can be specified using the container id or its associated name. Additionally it is possible to invoke run with an image name:

udocker run mytensor bash

---

5. run: II

Now you are inside the container (apparently as root), you might as well try out:

root@pcdavid:~# python
Python 3.8.10 (default, Nov 26 2021, 20:14:08) 
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf

Or:

udocker run mytensor cat /etc/lsb-release 

---

6. setup

With --execmode chooses an execution mode to define how a given container will be executed. The option --nvidia enables access to NVIDIA GPUs (only possible if they are available).

udocker setup --execmode=F1 mytensor
udocker ps -m  # confirm change of execution engine

---

7. rm

Delete a previously created container. Removes the entire directory tree extracted from the container image and associated metadata:

udocker rm mytensor

---

8. rmi

Delete a local container image previously pulled/loaded/imported:

udocker rmi tensorflow/tensorflow

---

End of Hands On part I