[DiGiTY]

I got a FAT32 drive and I want to essentially copy all the files and folders over to a EXT3 drive preserving their date and time. How can I accomplish this using Ubuntu/Linux or Windows?

[lotuseclat79]

Hi DiGiTY,

For the purpose of explanation, let's assume you have the latest Ubuntu 8.04 Live CD, i.e. Hardy Heron, and have booted it up in an environment where both of your disks - i.e. FAT32 (or vfat) and Linux (ext3) drives are available when you give the sudo fdisk -l command from a Terminal (command line) window.

First, you would need to become the root user and issue the command:
$ sudo -i
# fdisk -l
Then you would need to make the directory mount points for mounting both the FAT32 and the ext3 file systems:
# mkdir /mnt/fat32
# mkdir /mnt/ext3
Then you would need to mount each file system from the data output by the fdisk -l command, like the following (where you supply the partition device names to mount - i.e. sxxx and hxxx):
# mount -v -t ext3 /dev/sxxx /mnt/ext3
# mount -v -t vfat /dev/hxxx /mnt/fat32

One method I would definitely favor would be the following use of the tar command from the Linux Live CD environment, assuming Linux root user:

# (cd sourcedir; tar -cf - .) | (cd targetdir; tar -xf -)

The sourcedir would be the full path name on the FAT32 drive from the mount point, i.e. /mnt/fat32/... where ... is the name of the topmost directory you want to copy, and similarly for the targetdir - i.e. /mnt/ext3/...

Give this method a shot before you try anything else, as it is the easiest way to get what you want (caveat: if it works - and I think it will, but have not tried it).

-- Tom

[TerryNet]

In Linux the cp command has a 'preserve' option to preserve such things as timestamps.

Maybe I'm not understanding all the issues, but it seems like cp with the proper arguments would do the job.

[lotuseclat79]

Originally Posted by TerryNet

In Linux the cp command has a 'preserve' option to preserve such things as timestamps.

Maybe I'm not understanding all the issues, but it seems like cp with the proper arguments would do the job.

Hi TerryNet,

The cp command will work, but you have to give it the following arguments:
$ cp -rp ./* <destination>

The -r argument is necessary to recursively include directory contents.

-- Tom

[DiGiTY]

so what's the difference between using "# (cd sourcedir; tar -cf - .) | (cd targetdir; tar -xf -)" and "cp -rp ./* <destination>"? are the results different? is one faster than the other (I am moving 1 TB of data)?

does the tar option actually archiving the files to the destination or copying the actual files and folders?

[lotuseclat79]

Hi DiGiTY,

The difference is that the first form (tar) will work over a network whereas cp is local to your system - though you can either use rcp instead of cp over a network, and can certainly mount remote file systems over a network to use the cp command. Note: no tar archive is stored permanently for this operation - I assume you have 1 GB or more RAM to do the operation locally - is that correct? The less memory you have, the longer it will take to transfer 1TB, so make sure you have shutdown any unnecessary executables when you are performing the transfer of data - and you might just disconnect from the Internet if this is a local operation whether over a LAN or just within your computer system from disk to disk.

Use scp, which uses ssh, for a secure copy over a network.

Most, if not all Unix/Linux administrators would probably prefer using the tar command for a job a large as yours, though some might not. The results should be the same in any case (i.e. no file data lost moving the files to a differently formatted file system), though if I wanted an "exact" copy I would create a partition at least the size of the data you are moving onto another disk, and use the dd command to do the copy to a like fat32 formatted file system (assuming the Linux Live CD model as I stated). That way the disk segments are copied precisely and you can control the block size that is transmitted (since you are moving 1TB of data). See the man page for dd for more information.

The tar command will preserve everything and uses a pipe to do the job while the (r)cp command(s) need(s) the explicit parameter -p to preserve the file attributes - tar is basically more flexible and reliable between file systems. If you forget the -p or -r parameter using any of the cp commands - you don't get what you wanted.

The thing that would concern me the most in your case is what if the transfer is interrupted (say due to a power outage) when you attempt the copy? Is your system backed up with UPS - uninterruptable power supply? Just make sure you have enough space at the destination storage, and hopefully all goes well regarding continuous power during the operation. It would be nice to use a backup system or method that can at least take a snapshot, and resume the operation if it is interrupted since your data set is so large (1TB).

-- Tom

[DotHQ]

I would suspect the tar would use extra overhead to first tar the data and then to extract that data from the temp tar file.
I would go with cp or rcp for this particular task. If not on a private network the scp should be used, but it will be slower for sure.

[DiGiTY]

well it's 900 GB of data i'm transferring from one drive to another. the fastest method would be ideal especially if it keeps timestamps in tact

[avisitor]

I'd assume cp is faster. A block level transfer is fastest, though, but not possible for this application.

You could benchmark it on, say 1GB of files, and see which is faster. I have a hunch it'll be cp, but the actual difference shouldn't be that great.

[DiGiTY]

"cp -rpufv ./ /media/DAISY" seems to be working just fine (the update option is used because I previously moved 400 GB of data over already for test purposes). so far the files AND folders timestamps are preserved