> I don't think there is a good way to copy that many hard links. All the
> tools to do it must build a table that ties the inodes to the files and
> search the table for a match for every file with a link count above 1.
> It just doesn't scale well. When I needed to move a similar size
> archive I started over and just kept the old drive around in case I
> needed to get to the older runs.
How about taking another approach - doing it piecemeal?
1. Copy the pool and cpool directories using cp, tar, rsync, or
whatever - there are no hard links within this tree.
2. For each pc/HOSTNAME directory, use rsync to copy both it
and the pool or cpool. Something like (untested)
rsync -aH pool cpool pc/host1 /new/location
This command will do very little writing, since it will
notice that it only needs to re-create links in the new
You might get better overall throughput by doing more than
one client directory at a time.
I haven't tried this, but I think it should work.