Installation¶

Server-side Installation¶

Prerequisites

Note

Installation of the server-side Atlas components are handled by Dessa for pilots hosted on Dessa's servers. You will not need to perform a server-side installation in this situation

Docker version >18.09 (Docker installation instructions)
Python >3.6
>5GB of free machine storage
The atlas_installer.py file
Ubuntu LTS 16.04 or higher

Procedure 1. Create a working directory for the Atlas installer

Copy the atlas_installer.py file into this directory.
Create and activate a Python >3.6 virtual environment using Conda or venv to minimize dependency issues.
Run the install script with python atlas_installer.py.

Client-side Installation¶

Prerequisites

Python >3.6
The atlas_installer.py file

Note: Atlas client-side components currently support Mac OSX, Linux & Windows 10.

For Linux and OSX

Create a new, empty directory where you will install Atlas.
Copy the atlas_installer.py file into this directory.
Create and activate a Python >3.6 virtual environment using Conda or venv to minimize dependency issues.
Run the install script with python atlas_installer.py -Ndpsc.

Navigate to the ~/.foundations/config/submission/ directory, and replace contents in the scheduler.config file with the following lines of code:

# scheduler.config
cache_config:
  end_point: /cache_end_point
container_config_root: /home/purefacts/.foundations/config/local_docker_scheduler/worker_config
job_deployment_env: local_docker_scheduler_plugin
job_results_root: /home/purefacts/.foundations/job_data
scheduler_url: http://pilot-purefacts.eastus.cloudapp.azure.com:5000
working_dir_root: /home/purefacts/.foundations/local_docker_scheduler/work_dir

For Windows 10

Create a new, empty directory where you will install Atlas.
Copy the atlas_installer.py file into this directory.
Create and activate a Python >3.6 virtual environment using Conda or venv to minimize dependency issues.
Install the following python package into the environment created in step 3: pip install pypiwin32==223
Run the install script with python atlas_installer.py -NdpscC

Navigate to the ~/.foundations/config/submission/ directory, and replace contents in the scheduler.config file with the following lines of code:

# scheduler.config
cache_config:
  end_point: /cache_end_point
container_config_root: /home/purefacts/.foundations/config/local_docker_scheduler/worker_config
job_deployment_env: local_docker_scheduler_plugin
job_results_root: /home/purefacts/.foundations/job_data
scheduler_url: http://pilot-purefacts.eastus.cloudapp.azure.com:5000
working_dir_root: /home/purefacts/.foundations/local_docker_scheduler/work_dir

Tip

Running python atlas_installer.py --help will give you troubleshooting advice if the script isn't working as expected.

Server Start-up¶

Note

Start-up of the server-side Atlas components are handled by Dessa for pilots hosted on Dessa's servers. You will not need to perform a server-side start-up in this situation

After completing the server-side installation, you can do the following to start Atlas:

Validate that you are in the same Python environment that was used to run the installation script.
Run atlas-server start.

Tip

You can also start Atlas with GPU support by running atlas-server start -g. This will allow Atlas to use all CUDA-enabled GPUs on your system.

Success

To validate that the Atlas GUI is working, go to the GUI. This is the preconfigured location for your Atlas GUI.

Hello Atlas¶

After completing the Server Start-up, try the following to launch your first Atlas job:

Navigate to where you'd like to create your Atlas project directory on your client machine.
Ensure that you are in the environment that was used during installation.
Run foundations init hello-atlas to create a sample project called hello-atlas.
Navigate to the newly created hello-atlas directory.
Run the sample python main.py code provided.
Head to your GUI to see the results!