Well it has been a challenge like anything new tends to be. It took many months of work to get the approvals to purchase the software and get R "through the front door". Once we purchased the software it has taken almost three months to get a Linux server set up and the RStudio software installed. There is still more to do before the server can be rolled out to other users. I am using it for my work and it is sweet. In our organization IT is siloed in it's own building far removed from where I work. Progress seemed to go fastest when I was able to work in the IT building directly with the Linux admin. In my experience everyone in IT is very nice and helpful, but it really helps if you have face to face contact and can spend some time working together on the setup.
Authentication was by far the biggest technological hurtle and a big reason why the setup has taken so long. This is an area I know almost nothing about, and I would prefer not to get into the weeds of it. We used PAM with Active Directory for both RStudio Server Pro and RSConnect. RStudio was very helpful providing some support on the phone to help us work through issues. Everything works now but there are still some lingering issues to resolve that I don't fully understand. I do think it will be worth going through the pain of integrating the RStudio software with our existing systems. Using PAM allows database authentication from within the RStudio server IDE to use be effortless which is great.
The actual process of setting up the server from my point of view was just installing a bunch of dependencies on the Linux server, building a couple versions of R from source, and then installing the RStudio software. All of that went pretty smoothly. I had practiced building R form source a few times using Docker on my laptop as others have mentioned on this thread.
The most common hiccup seemed to be getting some weird error, googling the error, and installing some Linux package that fixed the error. I actually do not have rights to install Linux packages. The Linux admin installed any packages I needed for me and gave me rights to run specific commands as root using sudo. Also I should mention that a couple times things were not working one day and then started working the next without us doing anything. Don't ask my why.
I'm still trying to figure out package management. I had been thinking I would just install all the packages people want to use in the global shared library on the server but @cole strongly recommended against this approach, and I plan to heed his advice. The best approach for controlling packages on the system seems to be to set up an internal CRAN mirror and there are a few options I am exploring.
Takaways from my experience so far:
- Expect the setup to take a while and to run into issues (hopefully you will be pleasantly surprised)
- Befriend the IT folks/Linux admins and try to get some face-to-face time to work on the server together
- It can be tough continuing to do your normal day job while also trying to get the R infrastructure working and a supportive manager is very helpful
All that said I am really looking forward to putting this software to use and doing some cool stuff with it. I am an R & RStudio superfan and believe that all the work to set up good R infrastructure will pay off.