chore: upgrade R and RStudio versions #218
Conversation
|
@gavin-k-lee I think all the pieces (i.e. dependent libraries) that needed new releases for this to work are now done. I did some tests in a different branch. Are you ok for me to transfer these changes here so that we can have images with newer R and Rstudio versions? |
|
@olevski yes sure we can transfer them or make a new PR and scrap this one. up to you, i don't mind. btw on an un-related note, was there any reason why we are building bioconductor-renku images as well? @rokroskar historical reason? |
gavin-k-lee
changed the title
|
@gavin-k-lee the bioc image was requested some time back - it takes a long time to build so it was nice to have a base image as a starting point. |
|
Also this is a good guide on the insanity that is R session environment variables. If I change only the See: https://rviews.rstudio.com/2017/04/19/r-for-enterprise-understanding-r-s-startup/ |
|
@olevski yes the RStudio environment setup is mind-boggling. It basically ignores any reasonable convention that you might expect 😭 🤦 |
|
Though I'm a bit confused why adding the paths to |
olevski
force-pushed
the
rocker-4.1.2-rstudio-1.4.1717-3
branch
from
7b018d4 to
e524791
Compare
|
btw I think the bioc builds might fail since the R and RBioc versions aren't corresponding exactly. |
olevski
force-pushed
the
rocker-4.1.2-rstudio-1.4.1717-3
branch
from
e524791 to
0422a7e
Compare
|
@gavin-k-lee the bioc builds were failing because of mistakes (that I just introduced) in the RSTUDIO_VERSION environment variable not being properly defined in the dockerfile. They should all pass now. |
olevski
force-pushed
the
rocker-4.1.2-rstudio-1.4.1717-3
branch
7 times, most recently
from
645ede5 to
463e898
Compare
olevski
force-pushed
the
rocker-4.1.2-rstudio-1.4.1717-3
branch
from
463e898 to
ade8dce
Compare
olevski
changed the title
|
Ok this is ready to go now. Here is a test project with bioconductor 13: https://dev.renku.ch/projects/tasko.olevski/test-image-upgrade-r-1 |
|
Here is a project with just R 4.1.0: https://dev.renku.ch/projects/tasko.olevski/test-r-project-1 |
|
Both of the test projects work! |
|
nice! I just tried both projects and when opening in an external tab it worked great and I tested the plotting (which was the most prominent problem before). In the iframe RStudio didn't load if i didn't touch it, but if i refreshed it took a long while to load - not sure if you experience that all? Maybe it is a resource issue? |
|
@gavin-k-lee I have seen the loading issues in the iframe you mention on other versions (in production) too. I am pretty sure it is related to resource usage. I never saw it in test but then I made sure to request 1cpu with the sessions - not sure if that helped really. I think the cpu should not be limited in k8s. Either way I think this is a problem that shows up sometimes and is not a result of the new images. |
|
On another note... I def lost some sanity just trying to get these images to build btw. I did notice that building the dockerfiles worked differently on my mac locally and in github CI. But then after I switched to my personal latptop (which runs linux) I could finally get the same outcomes I see in the CI workflows locally and eventually fix them. Just a PSA if some else is having such weird issues. |
|
yeah it seems like you really need to build them with linux to have them equivalent, which is annoying. @SwissDataScienceCenter/renku-ui-maintainers would we be able to get some help to understand what's going on with the iframes? I tried the highest cpu and ram on dev and the iframe didn't load (but the new tab worked as expected). |
|
I think we can also retire some of the 4.0.X images and bioc images. We could go with: R 4.0.5 came out in March 2021 and Bioc 3.12 in October 2020. Thoughts? |
|
I just tried both the test-image-upgrade and the test-r-project above, and I did not see anything unusual in the iframe experience. Both projects took 40-45s to start, but I had the spinner/bouncer and then they transitioned seamlessly to RStudio once the pods were up. It seems that the problem is intermittent. If you could note the time (UTC) when you next have issues, we can see if there is anything in the logs that sheds light on what is going on. |
|
@rokroskar I tested the latest version of the R images with the If you or anyone else is interested here are the projects: |
We have been out-of-date with R/RStudio for a while now (#154 & #160) due to external dependencies. Since these external dependencies have been resolved, we can build newer versions based on Jupyter proxies.