Anaconda python distribution is very convenient to deploy scientific computing env (SCE) and switch python versions as you want. By default, the installation will locate python into ~/anaconda and the SCE can only benefit the local user.

But what I need is to provide a complete SCE wit Anaconda while masking the system-wide python version, because my cluster is running Spark and provides services for multiple users in our team. Is it possible with current Anaconda version?


The installer lets you install anywhere. You can install it to a global location, like /opt/anaconda.

Add the anaconda PATH to /etc/profile:

for anaconda 2 :


for anaconda 3 :


and then :

source /etc/profile

Anaconda Cluster from Continuum that addresses these issues. Check out

You definitely need to install to a path which is accessible from all users. /opt/YOUR_CONDA_DISTRIB_NAME is a good candidate.

However you also have to add conda path which automatically happens when you add the following shell script to other users’ .bashrc files. You can find the shell script in the root user’s .bashrc file. For the record, I bring it here:

# >>> conda initialize >>>
# !! Contents within this block are managed by 'conda init' !!
__conda_setup="$('/opt/YOUR_CONDA_DISTRIB_NAME/bin/conda' 'shell.bash' 'hook' 2> /dev/null)"
if [ $? -eq 0 ]; then
    eval "$__conda_setup"
    if [ -f "/opt/YOUR_CONDA_DISTRIB_NAME/etc/profile.d/" ]; then
        . "/opt/YOUR_CONDA_DISTRIB_NAME/etc/profile.d/"
        export PATH="/opt/YOUR_CONDA_DISTRIB_NAME/bin:$PATH"
unset __conda_setup
# <<< conda initialize <<<

P.S. a few common names which Anaconda commonly uses for YOUR_CONDA_DISTRIB_NAME:

  • anaconda
  • anaconda3
  • miniconda
  • miniconda3