🚀 go-pugleaf

bpalmer@pcug.org.au (Bryan Palmer) writes:

>It's not the size of the filesystem which vexes me (I gather Linux
>allows for filesystems in Terabytes), it is the size of individual
>files. I am looking to do some computational work on a data set of
>about 6Gb in size. I don't want to partion or pack the file; and as it
>stands 2.0.xx Linux does not allow files beyond 2Gb.

I've heard suggested before to just use a raw partition instead of a
file.  For example, just go in as root and make /dev/sda3 owned by
you, then make a symlink to /dev/sda3 so your program will use it as
its data file.  I haven't heard how well this works in practice.

That wouldn't have helped me the other day, though.  I had a program
running like so:
	my_program > debug.log
Unfortunately, debug.log grew very fast, exceeding the size of the
filesystem.  However, by periodically doing:
	echo >debug.log
I would truncate the file, but the original program just kept writing
it's output.  The file size reached 1.5G before it died, but it only
took up a hundred megs or so.  Aren't holes wonderful?  Needless to
say, I was getting quite worried about the 2G file limit, even though
my filesystem was less than 1G.

--PC
--
DOE SHP @ NMFECC/LLNL (NERSC) ID 87