I have an SFTP server where clients are constantly uploading large files. On a periodic basis, I want to copy all complete (fully uploaded) files to another machine for processing. I don't want to copy a file that is actively being written to. Is there a way to accomplish that? I am currently using rsync but I am open to switching to something else.


To check if a file is currently open (if a file is currently written is for sure open by some process) the standard way is to use lsof:

if lsof /your/file > /dev/null; then echo "file currently open"; fi

You can use this snippet to filter find results for only not opened files and use them to feed rsync:

find . -type f -exec sh -c 'if ! lsof `readlink -f {}` > /dev/null; then echo `basename {}`; fi' \; | tr '\n' '\0' | rsync -avz --from0 --files-from=- ./ user@host:destination/

Some notes:

  • readlink -f is needed to have full path of a file, lsof accept only full path
  • tr '\n' '\0' emulate find -print0
  • Where I specify the path to rsync ?
    – Freedo
    Apr 4 '17 at 8:38
  • The last part of the command: user@host:destination to rsync to a remote path or simply destination to rsync to a local path.
    – lgaggini
    Apr 5 '17 at 4:52

One challenge here is to determine whether the files are still begin written to. There is no perfect way to do this. I think the best you can do is to simply check the last-modified timestamp on the files, and only copy those files that have not been modified for a few minutes.

rsync by itself cannot do this, but you can combine it with the find-command:

cd /path/to/directory/with/files
find ./ -type f -mmin +5 -print0 | rsync --archive --verbose --from0 --files-from=- ./ yourotherserver:targetdir/

To break down this command, it does two things:

  1. It uses find ./ -type f -mmin +5 -print0 to identify all files that haven't been modified for at least 5 minutes.
  2. It then feeds this list into rsync using the --from0 and --files-from parameters. This will make rsync only consider those files that find has identified.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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