Hey all, I have 400 GBs of data (movies, music, code, etc) on an NTFS drive to move to a new 1.5 TB mirror (software) raid all on the same computer. So I was wondering what the best way to go about this would be? Is 'cp' enough? Will ntfs-3g barf in the process?

Also, I haven't setup the 1.5 TB Raid yet and was wondering if everyone thinks I should format it as XFS or ext3 or what? I'll have LOTS of large files on it and it will be used as a samba share as well.

cp should be enough, though you probably will want to setup some ownership and permissions after that. There's no way you could "import" those from the ntfs because the windows file systems do not have the necessary structures anyway.

But you can worry about that later. There are way to do mass-processing for that so you don't have to go clicking silly permissions boxes around.

cp should be ok, you could use rsync as well. Any tool that can copy files should be ok.

When people speak about BIG files for XFS they often do not know what a BIG file is in this context. A big file for XFS is more like in the range of TBs, or at least more than a couple of dozen GBs. Any fs will handle files of a few GBs just fine, you don't need XFS just for that. Id personally use ext4 if your kernel is recent enough (yes it's new, but it's stable and it's been merged in 2.6.28 in the stable kernel branch, so it's well tested and it's a reliable fs). Otherwise just use ext3 and tune it this way:

That should be fine for your purpose. And you can upgrade to ext4 at any given time without reformatting.

I don't have any particular thing against XFS, but I just don't advice it unless you can guarantee a non-interruptible power supply. It has demonstrated for me and many people not to be reliable when there are unwanted shutdowns.

Thanks man! I wasn't too worried about permissions as I was meaning to redo most of it anyways. And I'll probably go with ext3 over ext4 since I'm on a 2.6.21 kernel (debian).
Thanks again!

Do not use XFS on a desktop system. It is toooooooooo slooooooooooow :^).
ext 3/4 is a good choice, and reiserfs for /usr simply kicks asses.

That depends. If you don't like to reinstall every once in a while and you write to /usr a lot (update regularly) you are not going to like the degradation of performance of reiser 3.x that the fragmentation produces. It's also very heavy in cpu term for my tastes, overall if tail packing is enabled.

If I was to use ext4, would it be stable enough to run as a raid 5 (software) volume?

I haven't use it under raid5 myself. But I can't think of a reason why it would be any less stable than ext3, which is rock solid.

I've used ext4 without problems since .28 became stable, and by that time I had been already using ext4dev for quite some time without noticing any problem, though I only used that on non-critical tasks where I could spare the contents if something bad was to happen.

Of course, if it's a critical system you should still do your tests regardless of what I or someone else says here.