Installation

The tutorial will demonstrate how one can use the Midas framework, namely, the Pilot job abstractions in radical.pilot in order to easily parallelize a large number of calculations with the MDAnalysis library. First we need to install MDAnalysis and radical.pilot.

We will use the conda package manager and set up a virtual environment to keep the computational environment (named mdaenv) clean for the tutorial.

Note

You need to set up working environments on any machine where you want to launch jobs (e.g., a cluster head node, a workstation, or your laptop) and run jobs (i.e., typically the compute nodes of your cluster).

Installing MDAnalysis

MDAnalysis is full installable with conda (but if you want to, you can also install it directly from source 1).

conda

Install the anaconda or miniconda distribution (Python 2.7) by following the links and the instructions. Install Python 2.7.

For example, on Linux:

wget https://repo.continuum.io/miniconda/Miniconda2-latest-Linux-x86_64.sh
bash Miniconda2-latest-Linux-x86_64.sh

Accept the modification of your start-up files; if you later do not want to use the anaconda distribution, remove a line like

export PATH=/home/USERNAME/miniconda2/bin:$PATH

from your ~/.bashrc file.

MDAnalysis and dependencies

The current MDAnalysis 0.16.x only fully supports Python 2.7 so we will create a Python 2.7 environment. By installing the MDAnalysis and MDAnalysisTests packages we automatically install all dependencies as well (such as numpy and scipy as well as various plotting and data processing libraries that will be used for plotting the results) 2:

conda config --add channels conda-forge
conda update --yes conda

conda create --yes -n mdaenv python=2.7
conda install --yes -n mdaenv mdanalysis mdanalysistests

Activate the installation (has to be done in every shell):

source activate mdaenv

Check success 3:

(mdaenv) $ python -c 'import MDAnalysis as mda; print(mda.__version__)'
0.16.0

(This should show a version of 0.16.0 or higher.)

Installing Radical Pilot

We follow the radical.pilot installation instructions — see there for details. The following are all the required steps but they leave out background details or trouble shooting tips. If something is not working, please see radical.pilot installation instructions or, if you are at a workshop, ask an instructor.

Install in the same virtual environment 4 as MDAnalysis 3:

(mdaenv) $ pip install radical.pilot

Check success 3:

(mdaenv) $ radicalpilot-version
0.45.1

You also need a MongoDB instance and set the environment variable RADICAL_PILOT_DBURL either with full username and password information

export RADICAL_PILOT_DBURL="mongodb://user:pass@host:port/dbname"

or, for open installations, just

export RADICAL_PILOT_DBURL="mongodb://host:port/dbname"

Note that you can set up a free MongoDB at http://mlab.com.

If you are doing this tutorial as part of a workshop, the organizers will provide you with a working MongoDB installation and instruct you what to put into RADICAL_PILOT_DBURL.

Finally, set up password-less ssh (or password-less gsissh with certificates).

Setting up the environment

Although you don’t need to install the software every time, you do need to ensure that you always work in the configured virtual environment mdaenv 4 and that you set set the environment variable for the MongoDB:

source activate mdaenv
export RADICAL_PILOT_DBURL="mongodb://user:pass@host:port/dbname"

Footnotes

1

If for any reason you want to perform a MDAnalysis source installation, follow these steps to install MDAnalysis devel manually.

The following assumes that you are working in a virtual environment or that you do a user installation (pip install --user but in the following the --user flag is omitted). It also assumes that you have a complete compiler tool chain installed.

First get pre-requisites:

pip install numpy cython

Get the latest source code from the source code repository (you need to have git installed):

git clone --depth=50 https://github.com/MDAnalysis/mdanalysis.git
cd mdanalysis

Check that you are on the development branch:

git branch

Should show:

* develop

Install development version:

pip install package/
pip install testsuite/

Check as above that you can import MDAnalysis (see main text):

$ python -c 'import MDAnalysis as mda; print(mda.__version__)'
0.16.1-dev

This should show a release number with the -dev suffix to indicate a development version.

2

Optionally, for a full feature environment you may also install clustalw2 (for sequence alignments with MDAnalysis.analysis.align.fasta2select()):

conda install --yes -n mdaenv -c biobuilds --yes clustalw=2.1

This is not required for this tutorial but if you start using MDAnalysis for other projects, you might want to be able to use all features without having to think about installing additional optional dependencies later.

3(1,2,3)

In the following, the shell’s prompt is shown as (mdaenv) $ and should not be typed. It is supposed to remind you that you must be in the virtual environment 4. Only type what follows after the prompt. If the commands give any output, it is shown on the lines following the input.

4(1,2,3)

Do not forget to activate the mdaenv environment whenever you open a new terminal:

source activate mdaenv

Otherwise, you will probably find that scripts cannot find MDAnalysis or radical.pilot.