Re: newfs limits? 10TB filesystem max?

On Sat, Feb 19, 2005 at 06:40:48PM -0600, Eric Anderson wrote:
> Bernd Walter wrote:
> >Creating sparse files, e.g. by using dd, is prety much unix basics.
> >And via md(4) you can get a disk type device from a file.
> 
> Sorry - I understand how to make a file with dd, but 5000TB filesystem 
> means to me someone has 5PB of space to put the filesystem on.. I had not 
> heard anyone call a file a 'sparse file' with regards to dd before this, 
> and the man page info for dd and sparse isn't all that telling. 

dd is just a tool to write a single block at a fileoffset.
You can also do with truncate - I usually use dd because truncate is
not avalable e.g. under Solaris.
It an UFS feature to not allocate space for file ranges that have never
been writen.
A file without continuous space allocation is called a sparse file.

> >testdisk=/tmp/testdisk
> >dd if=/dev/zero bs=512 count=1 oseek=2m of=${testdisk}
> >mdev=`mdconfig -a -t vnode -f ${testdisk}`
> >
> >I don't know if md(4) works with such large disks, but it's very likely
> >that is does.
> 
> I see that running the command gives a 1GB file, that takes very little 
> disk space.  I must have missed this option in the dd man pages, or never 
> looked for it.  

However - you need your filesystem setup to support such large files.
That is large fragments to allow large allocation chains with big
fragments each.
In my case I was limited to 128T and since I don't want to newfs the
backing filesystem that's my limit for now without concatenating
multiple of them.

-- 
B.Walter                   BWCT                http://www.bwct.de
bernd_at_bwct.de                                  info_at_bwct.de