When MRAN is down, the build fails

[sjspielman]

Opening issue to note that since Friday 3/18/22, builds have been recurrently failing during Data Download. Here is an example of one of those fails.

It appears to be an issue with MRAN -

Error: (converted from warning) unable to access index for repository https://mran.microsoft.com/snapshot/2019-07-05/src/contrib:
cannot open URL 'https://mran.microsoft.com/snapshot/2019-07-05/src/contrib/PACKAGES'

Tagging @jashapiro @jaclyn-taroni for ideas here

[jashapiro]

Looks like MRAN is just plain down:
No dates on https://mran.microsoft.com/timemachine work.

[sjspielman]

I played around trying to get a snapshot on my computer too. To me this is pretty clearly down...

Error: Cannot install packages:
* cran::stringr: Can't find package called cran::stringr.
* cran::rprojroot: Can't find package called cran::rprojroot.
* cran::readr: Can't find package called cran::readr.
* cran::dplyr: Can't find package called cran::dplyr.
* cran::ShatterSeek: Can't find package called cran::ShatterSeek.
* cran::ggplot2: Can't find package called cran::ggplot2.
* cran::ggpubr: Can't find package called cran::ggpubr.
* cran::RColorBrewer: Can't find package called cran::RColorBrewer.
* cran::tidyr: Can't find package called cran::tidyr.
* cran::forcats: Can't find package called cran::forcats.
* cran::rmarkdown: Can't find package called cran::rmarkdown.

[sjspielman]

Sort of concerningly, the MRAN folks are MIA. See here as an example. There are several other very recently opened issues.

[jashapiro]

We might be able to switch the repository for the image to rstudio package manager: that repository does seem to go back far enough, so we could use the closest date: https://packagemanager.rstudio.com/cran/2019-07-07.

We'd have to investigate a bit how rocker stores the CRAN mirror (I think it is in /usr/local/lib/R/etc/Rprofile.site based on https://github.com/rocker-org/rocker-versioned/blob/1920e7cfc757bad02d041a0bddec1a18b1ebc4c1/r-ver/3.6.0.Dockerfile#L118)

I'm happy to implement/test this if it is the direction we want to go.

[jashapiro]

I played around with this, and I can confirm that with the following change, the Docker image builds using RSPM instead of MRAN (but I can not confirm that everything is identical):
jashapiro@b6836f4

However, MRAN seems to be back up, so this may be moot, for now.

[sjspielman]

MRAN is still back up, so for now this is moot. If this situation comes back to haunt us we could consider a different solution!
I'll keep this open for now as discussion as needed.

[jharenza]

This is haunting me over in https://github.com/PediatricOpenTargets/OpenPedCan-analysis at the moment, so following this.