Cloudera environment variables are not accessible in Rstudio image

[quinten-goens]

To whom it may concern, I noted that in CML the RStudio image does not provide you access to the environment variables set in Cloudera. Nor does the Git GUI of RStudio work (which might be related).

[carlsonp]

Can confirm this is still broken with docker.repository.cloudera.com/cloudera/cdsw/ml-runtime-workbench-python3.7-standard:2023.05.2-b7 as the base image. Based on looking at the Dockerfile, it may have something to do with permissions and/or contents of /etc/R/Renviron.site.

[quinten-goens]

@carlsonp thanks for looking into this a little! When doing some research, I came to some suggestions to solve this issue. But looking at lines 57, 58 and 60 in the rstudio-cml image I saw this fix was already added. It might have worked at some point in time.

I am referencing the above so perhaps someone else with a better understanding of RStudio/Cloudera environments might see the problem. I still haven't found a solution unfortunately..

[quinten-goens]

I think line 57 might be failing perhaps.

env | grep -v ^SPARK_DIST_CLASSPATH >> /usr/local/lib/R/etc/Renviron.site

On my live container, I saw there's no folder /usr/local/lib/R/etc/. Only /usr/local/lib/R/ exists. So when you would execute the above line, it will fail because /etc/ does not exist. Could it be as easy as adding the following?

mkdir /usr/local/lib/R/etc/
env | grep -v ^SPARK_DIST_CLASSPATH >> /usr/local/lib/R/etc/Renviron.site

I cannot test it as I am not a CML admin.