R-Studio Server without local R-System Installation.....

Hello all,
I have a general question regarding the installation of the R-Studio Server, if this is done under your own responsibility.

I would very much like to set up the R-Studio server separately and use my "CLOUD - Wide" R for this, so that I don't have to support the R x times. Are there any restrictions here that speak against such a solution?

Has anyone already set this up so that you can even use an HPC in the background?

Kind regards from Berlin....

Z. Matthias

Hello,

Installation of the Open Source RStudio Server is done under your own responsibility indeed. For the professional products, e.g. RStudio Workbench, additional options exist to support installation and with regards to functionality, however.

With the pro-products you would gain access to the RSW Launcher amongst many other features. With the Launcher you can automatically launch your R session either locally or remotely on a SLURM based HPC compute node or on a Kubernetes cluster out-of-the-box.

I am not aware of any restriction for using your "CLOUD - Wide" R with Rstudio Server. You simply need to make it available on the RStudio Server and ensure that it still works (e.g. if mounting the R installation tree via NFS, you need to make sure that all OS dependencies are met on the RStudio server). If you can be more specific on how you want to use your "CLOUD - Wide" R, I'd be happy to discuss more.

If using OS RStudio Server you still can utilise various R packages that support remote submission to a HPC cluster. One of the more elegant and performant versions I have come across is the clustermq package. With a bit of tweaking you could use a passwordless ssh connection from your RStudio Server to the HPC and from there you could automatically get the jobs submitted to the HPC cluster, all via a concatenation of clustermq's ssh connector and the SLURM/LSF/... template (with clustermq you would not even be limited to have a SLURM cluster as it supports quite a few schedulers. Once set up you only could use the Q() function and the code chunk encapsulated in this function call will be remote executed.

Hello Michael,

that sounds very good and also very exciting. I think I will first make my experiences with the RStudio server and the clustermq package.

Well, I can make the package available to my web server via SSHFS so that I can also work "locally" with R. The idea of "clustermq" is very appealing, I will try it out afterwards. And yes, I run a small HPC cluster with SLURM (in the lab and also outside), hence my question here.

I will test this and if I have any questions, I will get back to you. Until then, thank you very much for the great help.