[BioC] writing of /usr/lib/R

[Liaw, Andy]

> From: lgautier at altern.org
> 
> > Hi folks,
> >
> > Thank you all for your cool ideas! And I'll make sure I 
> send questions
> > like this to the proper list next time.
> >
> > Actually I've discovered the R_LIBS env var in the FAQ, 5 minutes
> > after I hit the Send button ;-)
> >
> > Also, it's worth to note that we have luxury here of running
> > distributed SAN-attached file system accessible by all the nodes of
> > the cluster. It's General Parallel File System (GPFS) from IBM. So I
> > was thinking why not utilize that? I can copy /usr/lib/R to GPFS and
> > then symlink from all the nodes to it. That will also serve the
> > purpose of having all the nodes work against exactly the same
> > /usr/lib/R.
> >
> > On the other hand I'm cautious about making /usr/lib/R 
> shared because
> > of the race condition it can cause. Because what will 
> happen  if, for
> > example, these two users are running different R scripts on two
> > different nodes of the cluster and their R processes both try to
> > install/upgrade packages in the shared directory? Won't they clash?
> 
> There is a lock file during package install/removal, if I remember
> correctly. Otherwise the last one to install/remove wins, I believe...

It's when the package is in use, I believe, so that a package cannot be
overwritten when someone is using it.

Andy

 
> > I have presently only 2 users (power users I must admit) using R and
> > their number is not expected to grow between two and maybe several
> > more.
> >
> > --
> > Warm regards,
> > Michael Green
> >
> > _______________________________________________
> > Bioconductor mailing list
> > Bioconductor at stat.math.ethz.ch
> > https://stat.ethz.ch/mailman/listinfo/bioconductor
> >
> 
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> 
>