OAR 3

Status: alpha-dev4

Released date: Mar 9th 2020

OAR 3 a new version with better code architecture and new internal structure for scheduling.

• Same design architecture as OAR2 (based on high level components)
• Low source line number of code, under 30K (core under 20kloc)
• New more scalable scheduler (based on new data structures)
• Python 3.7
• Better development tools
• Unit test

• Alpha version (3.0.0dev4): Mar 9th 2020
• Pre-Alpha version (3.0.0dev3): Nov 12th 2018
  • Version to evaluate robustness of core functions
  • UI needs some cleaning
• Beta version: 2Q20 (current target)
• Production version (3.0.x): 2020

Note: Delays in previous roadmap are due to works around Batsim a simulator of infrastructures.

Test Alpha version uses Oardocker

• Install oardocker

 pip install oar-docker

• Second create Debian 10 based images and install 3.0.0.dev4 version oar.

 mkdir oar3 && cd oar3
 oardocker init -e buster build
 git clone --depth 1 --branch 3.0.0.dev4 https://github.com/oar-team/oar3.git
 oardocker install ./oar3

• Third launch containers, connect and submit an interactive job

 oardocker start -n 3
 oardocker connect frontend
 oarsub -I

Only for development and the use of Sqlite by default (no need of PostgreSQL), you can follow the following steps:

• Clone the master repository

 
git clone https://github.com/oar-team/oar3.git

• Prepare python environment based on 3.6 version and activate it

 
virtualenv -p python3.6 env3.6
source env3.6/bin/activate

• Install all dependencies

 
cd oar3 
pip install -e '.[dev]'

• Launch tests with coverage, html result is located at oar3/htmlcov/index.html

pytest --cov-report=html --cov-report term --cov=oar

• All core features are written
• Scheduling part ready at 90% (fully operational)
• Need OAR2 installation procedure to function
• Incomplete CLI, few options available
• Not all design/architecture choices are fixed
• Nodes energy saving unfinished
• Pingchecker unavailable
• SendMail to user unavailable
• Test coverage 70%-80%

• Standalone procedure installation
• Complete CLI commands
• Test coverage 80%-90%

• Version usable by our early adopters
• Pre-production version
• Documentation
• Test coverage >90%