Python and R with Anaconda

To support the diverse python and R workflows and high levels of customization Research Computing users require, Anaconda is installed on the CURC system. Anaconda uses the conda package manager to easily install software and associated packages, and supports python, R, and many other applications. The following documentation describes how to activate the CURC Anaconda distribution and our default environments, as well as how to create and activate your own custom Anaconda environments. Additional documentation on CURC OpenOnDemand is available for users desiring to interact with their custom environments via Jupyter notebooks.

Configuring conda with .condarc

The conda package manager allows modification of default settings to be done through a text file known as the .condarc. This file exists within a user’s /home/$USER/ directory and can be quickly be accessed using the file’s full path at ~/.condarc.

Your /home/$USER directory is small – only 2 GB. By default, conda puts all package source code and environments in your /home/$USER directory , and it will quickly become full. The steps here modify the conda configration file to change the default locations for packages and environments to your larger /projects/$USER directory.

Open your .condarc file in your favorite text editor (e.g., nano, vim):

Note: this file may not exist yet – if not, just create a new file with this name; you can open or create file with the following command
[johndoe@shas0137]$ nano ~/.condarc

…and paste the following four lines:

  - /projects/$USER/.conda_pkgs
  - /projects/$USER/software/anaconda/envs
Note: CSU and XSEDE/ACCESS users may need to use a custom $USER variable because the @ symbol in the usernames can occasionally be misinterpreted by environments that employ PERL. Directions to set up a custom user variable can be found at our CSU and XSEDE username documentation.

…then save and exit the file. You won’t need to perform this step again – it’s permanent unless you modify .condarc later.

The .condarc file provides a variety of settings that can be detailed to speed up your workflows. For more information on .condarc, check out the Anaconda documentation.

Using the CURC Anaconda environment

Follow these steps from a Research Computing terminal session on an Alpine acompile node or within a Alpine/Blanca batch or interactive job.

Activate the CURC Anaconda environment:

Run the following command to load the base Anaconda software:

[johndoe@shas0137 ~]$ module load anaconda
Note: The command above activates the base envioronment for python3, which as of 2020 is the only supported python standard. For users requiring legacy python2, you can still use conda to create a custom environment with the python2.X version of your choice (we provide an example of how to do this below).

You will know that you have properly activated the environment because you should see (base) in front of your prompt. For example:

(base) [johndoe@shas0137 ~]$

Using Conda:

Now that you have activated the base conda environment, you can use conda for python and R! There are two ways forward, depending on your needs. You can:

1. Use one of CURC’s pre-installed environments.

  • Pros: You can begin using one of these immediately, and they contain mainy of the most widely used python and R packages.
  • Cons: These are root-owned environments, so you can not add additional packages.


2. Create your own custom environment(s).

  • Pros: You own these, so you can add packages as needed, control package versions, etc.
  • Cons: There really aren’t any cons, other than the time needed to create a custom environment (usually 5-30 minutes depending on the number of packages you install).

Both options are discussed below.

Using one of CURC’s pre-installed enviroments:

To use the CURC environment with OpenMP, run the following command with Anaconda initialized:

(base) [johndoe@c3cpu-a7-u19-1 ~]$ conda activate 

You will know that you have properly activated the environment because you should see (pyomp_2022) replace the (base) in front of your prompt.

To see the python packages available in the environment, you can type conda list.

Similarly, to use the CURC R distribution (R v3.6.0), run the following command with Anaconda initialized:

(base) [johndoe@c3cpu-a7-u19-1 ~]$ conda activate 

You will know that you have properly activated the environment because you should see (rstudio) in front of your prompt. To see the R packages available in the environment, you can type conda list. Now, you can use R as you normally would.

Because interactive development is more easily done locally, most CURC R users exclusively run R code within batch jobs. Should you need to use rstudio on top of R for interactive development on Alpine, you can login to our system with X11-forwarding (ssh -X) and initiate an rstudio session from within an interactive job.

Create your own custom environment:

Note: In the examples below the environment is created in /projects/$USER/software/anaconda/envs, which is specified under envs_dirs in your .condarc file. Environments can be installed in any user-writable location the user chooses; just add the path to ~/.condarc.

1. Initialize Anaconda if you haven’t already done so:

[johndoe@shas0137 ~]$ module load anaconda
(base) [johndoe@shas0137 ~]$ 

2. Create a custom environment:

Here we create a new environment called mycustomenv (you can call it anything you want!)

(base) [johndoe@shas0137 ~]$ conda create -n mycustomenv

If you want a specific version of python or R, you can modify the above commmand as follows (e.g.):

Python v2.7.16:

(base) [johndoe@shas0137 ~]$ conda create -n mycustomenv python==2.7.16

Python v3.6.8:

(base) [johndoe@shas0137 ~]$ conda create -n mycustomenv python==3.6.8

Latest version of R:

(base) [johndoe@shas0137 ~]$ conda create -n mycustomenv r-base

3. Activate your new environment:

(base) [johndoe@shas0137 ~]$ conda activate mycustomenv
(mycustomenv) [johndoe@shas0137 ~]$ 

If successful, your prompt will now be preceded with (mycustomenv).

4. Install needed packages in your new environment:

The best way to do this for python packages is to install everything you need with one command, because it forces conda to resolve package conflicts. For example:

(mycustomenv) [johndoe@shas0137 ~]$ conda install numpy scipy tensorflow

For R packages, it is easiest to start an R session and then install the packages as you normally would with “install.packages”. For example:

(mycustomenv) [johndoe@shas0137 ~]$ R

If you encounter a --- Please select a CRAN mirror for use in this session --- message, you can select a US mirror from the provided list or use the repos install flag:

>install.packages('RMySQL', repos='')

For more information on managing conda enviornments, check out Anaconda’s documentation here.

Basic conda commands to get you started:

Command Function
conda list List the packages currently installed in the environment
conda search <package> Searches the Anaconda package channel for a package named <pakage>
conda install <package> Installs a package named <package> to your currently loaded environment
conda uninstall <package> Uninstalls a package named <package> from your currently loaded environment
conda env list List the conda environments currently available
conda create <env> Creates a new anaconda enviornment named <env>
conda remove --name <env> --all Removes an environment named <env>
conda deactivate Deactivates current enviornment


If you are having trouble loading a package, you can use conda list or pip freeze to list the available packages and their version numbers in your current conda environment. Use conda install <package> to add a new package or conda install <package==version> for a specific verison; e.g., conda install numpy=1.16.2.

Sometimes conda environments can “break” if two packages in the environment require different versions of the same shared library. In these cases you try a couple of things.

  • Reinstall the packages all within the same install command (e.g., conda install <package1> <package2>). This forces conda to attempt to resolve shared library conflicts.
  • Create a new environment and reinstall the packages you need (preferably installing all with the same conda install command, rather than one-at-a-time, in order to resolve the conflicts).

Dbus Error

If you see a ‘dbus’ connection error when trying to connect via a virtual environment:

Could not connect to session bus: Failed to connect to socket /tmp/dbus-oBg2HbRfLi: Connection refused.

This is likely due to your ~/.bashrc configuration file auto-activating a conda environment with a problematic dbus package. You can resolve this issue by opening your ~/.bashrc with a text editor (ex. vim, nano) and commenting out the following lines (or any lines that add a conda environment to your $PATH):

Note: Commenting lines out instead of removing them will allow you to add them back in later if needed. These lines have been commented out using # preceeding each line.
# >>> conda initialize >>>
# !! Contents within this block are managed by 'conda init' !!
# __conda_setup="$('/curc/sw/anaconda3/2019.07/bin/conda' 'shell.bash' 'hook' 2> /dev/null)"
# if [ $? -eq 0 ]; then
#     eval "$__conda_setup"
# else
#     if [ -f "/curc/sw/anaconda3/2019.07/etc/profile.d/" ]; then
#         . "/curc/sw/anaconda3/2019.07/etc/profile.d/"
#     else
#         export PATH="/curc/sw/anaconda3/2019.07/bin:$PATH"
#     fi
# fi
# unset __conda_setup
# <<< conda initialize <<<

Keep in mind that doing this means conda is not automatically sourced by your ~/.bashrc so you will have to manually source the base conda envioronment with module load anaconda to activate the base environment.