Skip to content

Latest commit

 

History

History
136 lines (89 loc) · 3.91 KB

MAINTENANCE.md

File metadata and controls

136 lines (89 loc) · 3.91 KB
Raw

Maintenance

Notes for the maintainers.

Test locally

Go to your local clone of the repository and install it as a regular Python package

module purge && module load JupyterLab
pip install -e .

Also fetch the corresponding build of the container. If you used a feature branch on the repo (you should ;-)), the CI would have built your container so you only need to fetch it:

module purge && module load Apptainer
apptainer pull <SIFPATH> oras://ghcr.io/nesi/rstudio_on_nesi:<BRANCHNAME>

where

  • <BRANCHNAME> is the name of your git branch,
  • <SIFPATH> is the path where to save the .sif image file.

Make sure to configure RStudio-on-NeSI to use this container instead of the default one.

mkdir -p ~/.config/rstudio_on_nesi/
echo <SIFPATH> > ~/.config/rstudio_on_nesi/singularity_image_path

Then start a new JupyterLab session on NeSI JupyterHub, start RStudio and check that everything works as intended ;-).

Once your test is done, make sure to cleanup your environment:

  • uninstall the Python package
module purge && module load JupyterLab
pip uninstall rstudio-on-nesi
  • remove the file indicating the alternative container
rm ~/.config/rstudio_on_nesi/singularity_image_path

Release a new version

Variable for convenience

Setup variables for your local git repo and the new version, repeatedly used in the workflow.

WKDIR=<PATH_TO_SIF_IMAGE> (The cloned git repo direcotry)
VER=<VERSION_NUMBER> (eg v0.25.0)

Always test locally the package and associated container first.

Go to your local clone of the repository and make sure to be in the main branch with all the latest changes:

cd $WKDIR
git checkout main
git pull

Increase the version number in rstudio_on_nesi/__init__.py, using the semantic versioning scheme.

Then commit, push, tag and push the tag

git commit -av -m "version $VER"
git push
git tag $VER
git push --tags

where $VER is the new version number

Check that the corresponding CI build job runs to completion without any error.

Then fetch the corresponding container image:

module purge && module load Apptainer
apptainer pull rstudio_server_on_centos7__$VER.sif oras://ghcr.io/nesi/rstudio_on_nesi:$VER

Move the sif image to a readable place for nesi-apps-admin` and update the permissions.

cp $WKDIR/rstudio_server_on_centos7__$VER.sif /nesi/project/nesi99999/
chmod 775 /nesi/project/nesi99999/rstudio_server_on_centos7__$VER.sif

Switch to nesi-apps-admin to update to the default image: sudo -iu nesi-apps-admin, variable is not passed to nesi-apps-admin and needs to be manually included. Once copied the .sif can be removed as it still resides in your local repo.

cp /nesi/rstudio_server_on_centos7__<VERSION>.sif /opt/nesi/containers/rstudio-server/rstudio_server_on_centos7__<VERSION>.sif
rm /nesi/rstudio_server_on_centos7__<VERSION>.sif

TEST TEST TEST

Finally, request an update of the JupyterLab environment module to install the new version of the Python package.

Test without Jupyter

This is an alternative way to test the container if you really don't want to use JupyterLab integrated launcher.

Start the container from the command line:

export SIFPATH=<PATH_TO_SIF_IMAGE>
PASSWORD=your_secret_password rstudio_on_nesi/singularity_wrapper.bash $(pwd)/rstudio_on_nesi/singularity_runscript.bash 9999 localhost

where <PATH_TO_SIF_IMAGE> is the path to the container image.

Note we will connect directly to the rstudio server, bypassing the Nginx reverse-proxy which is useless in this context without Jupyter.

Check the messages in the terminal to find out the rstudio server port, for example:

rserver cmd: /usr/lib/rstudio-server/bin/rserver --www-port 40074 --auth-none 0 [...]

would correspond to port 40074.

Finally, start an ssh tunnel on this port and open your web browser at http://localhost:<REDIRECTED_PORT> to access rstudio.