Currently I can only copy a single .tar file. But how can I copy directories recursively with scp?

up vote 803 down vote accepted

Yup, use -r:

scp -rp sourcedirectory user@dest:/path
  • -r means recursive
  • -p preserves modification times, access times, and modes from the original file.

Note: This creates the sourcedirectory inside /path thus the files will be in /path/sourcedirectory

  • 7
    However bear in mind that this won't preserve symlinks. – CpnCrunch Jul 24 '15 at 0:30
  • 6
    Note that -pr (options in reversed order) won't copy the folders, bur rather their content to the target directory (apparently, the order of options matters). – pms Feb 3 '17 at 14:49
  • For some reason /path appears to be relative to the users $HOME -- how do I make it absolute without being root@? – Jonathan Aug 25 '17 at 21:50
  • Yeah, SCP like never puts the files where I want. Often when trying to replace a directory, I somehow end up with another directory inside it with the same name instead. I'm extra careful with this. – sudo Jan 24 at 3:00

While the previous answers are technically correct, you should also consider using rsync instead. rsync compares the data on the sending and receiving sides with a diff mechanism so it doesn't have to resend data that was already previously sent.

If you are going to copy something to a remote machine more than once, use rsync. Actually, it's good to use rsync every time because it has more controls for things like copying file permissions and ownership and excluding certain files or directories. In general:

$ rsync -av /local/dir/ server:/remote/dir/

will synchronize a local directory with a remote directory. If you run it a second time and the contents of the local directory haven't changed, no data will be transferred - much more efficient than running scp and copying everything every time.

Also, rsync allows you to recover from interrupted transfers very easily, unlike scp.

Finally, modern versions of rsync by default run over ssh, so if scp is already working, rsync should pretty much be a drop-in replacement.

  • 11
    "no data will be transferred" except, of course, the data required to determine exactly what has or has not changed. – thsutton Apr 29 '11 at 5:15
  • 13
    Ha yeah I glossed that over a bit obviously. We're not talking quantum entanglement here. – Phil Hollenback Apr 29 '11 at 6:01
  • 2
    it's so sad that osx, win7 and not even ubuntu uses something like rsync for their GUI yet. – cregox Apr 29 '11 at 10:24
  • 11
    dmourati I did answer his question. I told him a better way to do it. – Phil Hollenback Jan 29 '13 at 21:23
  • 2
    @raphnguyen: no, deleting files from the local directory will not delete them from the remote directory. You can turn this behavior on with the --delete rsync option. – Phil Hollenback Oct 26 '15 at 20:19

That is what the -r option is for. :)

See the scp man page for more info if needed.

Recursive Copy Option '-r' (lower case)

scp -r

Which I confuse with the regular local recursive copy option '-R' (upper case)

cp -R

The best way is to use rsync over SSH

rsync -a -essh /source/ user@dest-server:/dest/

rsync -a -essh user@source-server:/source/ /dest/

My favorites options are -Pazvessh --delete :

  • -a : archive mode (include a lot of default common options, including preserving symlinks)
  • -z : compress
  • -v : verbose : show files
  • -P : show progess as files done/remaining files
  • -e ssh : do rsync in ssh protocol
  • --delete : delete files in the destination that are not anymore in the source
  • All versions of rsync that I have used would use ssh by default, so -essh is unlikely to be needed. And the choice of command used to connect to the remote host is really unrelated to copying recursively. – kasperd Nov 5 '15 at 19:03

After looking for the recursive copy flag, and successfully used it thanks to this post, I would like to post just a suggestion.

If the case is that you are copying (recursively) a directory. Maybe if the files are sent compressed you could save time in the transfer

What I did in the end was:

local$ tar -czvf local.tar.gz directory/
local$ scp local.tar.gz user@remote:/directory
ssh user@remote
remote$ tar -xzvf local.tar.gz

Hope this helps

  • the file extension should be either .tar.gz or .tgz since the file is a gzipped tar archive (since the -z flag is used). – anthonybell Mar 26 at 22:07

You can recursively copy a directory into a compressed archive with this simple command:

ssh -p 22  'cd /parent/directory && tar zcvf - directory_to_copy' > /destination/on/your/machine/archive_name.tgz

For example, to copy contents of /var/log from to ~/logs.tgz you run:

ssh -p 22  'cd /var && tar zcvf - log' > ~/logs.tgz

You can also extract files on target system by using pipes. This command will copy contents of /var/log at to ~/destination/log on your system:

ssh -p 22  'cd /var && tar zcvf - log' | tar xzf - -C ~/destination

Though to mirror a directory, you probably should use rsync...

If you prefer to pass the user's password as a parameter rather than inputting it interactively, you can use sshpass (sudo apt-get install -y sshpass).


sshpass -p 'remote_password' scp -rp /src/folder myremoteusername@

Another (likely better for repeated use) option is to use NFS - check out nfs-kernel-server and how to setup NFS shares.

  • 2
    This is complete nonsense. There are endless cases where you would use scp but NFS is not an option. – Sven Aug 11 '16 at 23:24

Your Answer


By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.