RStudio version too old

Yes, you have to browse to /rstudio manually after the notebook has opened.

1 Like

That seems to work indeed!

There’s that rstudio warning/message at the top that doesn’t seem to affect a simple usage, as far as I can say.

Hello - any updates on this on the Renkulab side? The ableuler/renkulab-r:rstudio-fix image works, but I still have the issue on Renku. What changes do I need to do to benefit from that fix. I seem to have updated my Dockerfile, but without success.

Hello @lgatto, we are still waiting for jupyterhub to fix this issue in order to make rsession work properly in jupyterhub.

I create another image st714/test:bioc-fix with RStudio=1.4.1717 , R=4.1.0 , bioc=3.13 and an unreleased version of jupyterhub-rsession-proxy. This image works on RenkuLab.

You need to change default_url in renku.ini from /rstudio to /rstudio/auth-sign-in?appUrl=%2Fworked to make it work properly.

For details, you could check this project. Let me know if it works for you.

2 Likes

Thank you very much @tao.sun - I’m giving it a test at the moment and will report back here.

Hi again - sorry, I don’t seem to get it working, as I get prompted for a login and password, and rstudio and rstudio don’t work. Could it be due to the next line in renku.ini?

image = registry.renkulab.io/laurent.gatto/wsbim-bioinfo:33ff09b

Hi @lgatto, yes, I think it’s about this line. The session will start with this pinned image instead of the recent one.

I’m building a new image at the moment (it takes quite some time to build) with a clean renku.ici. Will let you know later. Thanks again for your help!

Grrr - now it throws me an error 500 : Internal Server Error. I must have an error somewhere, but can’t spot it. I might be better off starting from your project.

Hi @lgatto, i forked your project and found that the error came from the last line in Dockerfile:

RUN pip3 install jupyter-rsession-proxy==1.2.0 jupyter-server-proxy==1.5.0

which actually undoes some fix I did.

For simplicity, I also commented the installation of R packages in my fork, which imo has nothing to do with this 500 error.

1 Like

Indeed, all is working now. Thank you again so much for your help!

1 Like

Hello!

I have successfully forked this project and am able to open a session with RStudio 1.4 with it.

I am wondering how to copy this behavior to other projects? Which files should I copy?

Thank you very much

Hi Kene,

Thank you for following up on this! We are currently working very hard to release a suite of R >= 4.1.0 + RStudio >= 1.4 Docker images and corresponding project templates - they should be ready soon and rolled out to the usual project creation pages shortly.

If you requirea quick fix immediately, the files you would need to look at in that project are the Dockerfile and .renku/renku.ini.

For the Dockerfile: Instead of FROM renku/renkulab-r4.0.5... that you may have in your existing R project, you could try the one in that project, which started with FROM ableuler/wsbim-fix:2.

For .renku/renku.ini file you should copy the one used in the project (the second line of it looks like this: default_url = /rstudio/auth-sign-in?appUrl=%2Fworked).

Let us know if that works or you need some extra help in your case!

Gavin

2 Likes

Thank you very much Gavin.

I seem to still be having a bit of an issue with this. When I change the renku.ini file, starting a session leads me to an RStudio login page, which I am unable to get past.

Might I be doing something wrong?

(Here is the project I would like students to fork and be able to run with RStudio >1.4)

Hi Kene,

We have started to build some R/RStudio images that are recent, but they have not been released yet/fully tested. However, you can still try it. You will probably need to open the session in a new tab to make sure it works.

Steps:

  1. Dockerfile: Use the following ‘from’ statement: FROM renku/renkulab-r:4.1.0-846186f
  2. .renku/renku.ini: Remove the ‘hacky’ existing one and use this:
[renku "interactive"]
default_url = /rstudio

You will then need to let the image build and then you can try to open a session once it’s built. The RStudio version will be >1.4 (they have renamed them to calendar versioning now), with R 4.1.0.

Thanks for your patience!! Let me know how it goes - we are still testing those images.

1 Like

Hello - thank you for updating the Renku R infrastructure. This is really important for us R/Renku users.

Just a little suggestion though - I see you target R 4.1.0. The latest version is 4.1.2; while it’s only a patched upgrade, it might we worth (and hopefully easy) to go the that one.

More generally, and to be able to get contributions from the community, where can I find the information to do the same for, say, R-devel? We aren’t too far from the release of R 4.2 (March - April, I think).

Hi there - thanks for the suggestion! We are planning on expanding the suite to:

R: 4.1.0, 4.1.1, 4.1.2, devel; and
Bioc: 3.13, 3.14, devel,

since R 4.0.5 came out in March 2021 and Bioc 3.12 in October 2020. Would that be sufficient at this point in time?

1 Like

Thank you very much! This seems to work for me!

1 Like

For now, these version are great, but the next version of R, and thus Bioconductor will be released in March, and quickly become desirable. These would, for example, become critical for all courses starting thereafter (for instance in Sept 2022).

The situation is a but confusing because R has a yearly release schedule (in March/April) and Bioconductor is released every 6 month and Bioconductor versions match specific R version. Currently Bioc 3.14 (stable release) requires R 4.1.1 (or 4.1.2) and Bioc 3.15 (devel) requires R-devel. Once R 4.2 is released, Bioc 3.15 will be released as stable version and a new Bioc devel branch will be started. (See also here for more details on the R and Bioc version dependencies.)

This essentially means that it is not unlikely that a new R/Bioc Renku environment would be needed. The Bioconductor docker containers could be helpful here.

Hi @lgatto, we should be OK to easily add new R versions in the future. The underlying incompatibility between Jupyter and RStudio seems to have been resolved, which is why we are finally unblocked on this (thanks to @tolevski and @gavin-k-lee). As far as the bioc images go, we are already using the bioconductor docker images as the base for those builds, see here.

1 Like