[R] Huge differences in Ram Consumption with different versions of R on the same scripts

[Eric Berger]

Assuming the vignettes are the same in both cases, try going back to
the system where it worked and introduce a single change? i.e.
a) previous linux system and R 4.2
b) Previous R version and Oracle Linux 8

Repeat the test for a) and b) and if one works fine and the other
fails at least that will narrow down the search.

HTH,
Eric




On Mon, May 8, 2023 at 4:41 AM Robert Knight <bobby.knight using gmail.com> wrote:
>
> Hello R Help,
>
> I have some R vignettes that run fpp2 time series regressions.  I have run
> these for a long time using R and a 12 core computer system.  I ran each
> core using Linux to run a vignette on that core, so that all 12 could work
> concurrently.  With 48GB of ram, the ram never filled up. I ran these
> regressions for hours, one data set right after the other on each core.
> Recently, I switched to Oracle Linux 8 and R 4.2 Now, with the same
> scripts, and the same data, the ram fills up and R reserves 4.2GB per
> instance in some cases.  This results in all the ram being consumed and the
> swap space on the system activating constantly so that the performance is
> abysmal. It begins using 12Gb of swap space in addition to the ram
> consumed.  It bogs the system so bad that one can't even establish new
> terminal sessions.
>
> Is there a way to specify a maximum ram allowance in an R vignette?  If
> that’s not possible, what resources can you recommend to help identify the
> reason for the change in memory use?  Why would a different version of
> R/Linux use 2GB per instance, while another uses 4.4GB? What kind of
> troubleshooting or programming techniques should I research for this kind
> of concern?
> Robert Knight
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.