So if I had a cp -v operation fail, is the last file name it printed out the last successful file copy, or is it the failed partially copied file? If you had to ensure all files are copied correctly without overwriting anything, would deleting the last filename that was printed from the destination folder delete the partially copied file that the operation failed on?

  • taladar@sh.itjust.works
    link
    fedilink
    arrow-up
    27
    ·
    edit-2
    1 year ago

    Just use rsync -va (possibly with --chown if you want user/group to be different at the destination and with --delete if you want removed files to be deleted) to continue the copy operation, it automatically takes care of figuring out which files still need to be copied and which are already there.

    • DefederateLemmyMl@feddit.nl
      link
      fedilink
      English
      arrow-up
      30
      ·
      edit-2
      1 year ago

      Just use rsync -va

      NO STOP!

      The default quick check algorithm of rsync is not safe for this. It only checks filesize and modification time to determine if files are equal. After a b0rked copy, these are not to be trusted.

      You should add the -c flag so that files are properly checksummed, unfortunately if you have slow storage on either end, this often negates the speed advantage of rsync.

      For example, consider this example:

      mkdir source
      mkdir destination
      echo "hello" > source/file.txt
      echo "world" > destination/file.txt
      touch -r source/file.txt destination/file.txt
      rsync -avh source/ destination/
      cat source/file.txt
      cat destination/file.txt
      

      Contrary to what you might expect, the rsync command copies nothing and the output at the end will show:

      hello
      world
      

      If you change the rsync command in the example above to rsync -c -avh source/ destination/, it will work as expected.

      • taladar@sh.itjust.works
        link
        fedilink
        arrow-up
        1
        ·
        1 year ago

        True if the initial state is unknown but if you do your initial copy and all the later syncs with rsync it is not really necessary since rsync puts the partial files in a temporary location (there are same parameters to control the details of that too).

      • damium@programming.dev
        link
        fedilink
        English
        arrow-up
        1
        ·
        1 year ago

        My memory of the cp command is that attributes such as file times were transferred at the last step. I think this would make rsync safe in most situations where a system crash wasn’t involved.

  • FigMcLargeHuge@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    9
    ·
    1 year ago

    Run an md5sum command on the last file in each location and compare. That will tell you if one of them is different and answer your question.

  • yuli [she/her]@hexbear.net
    link
    fedilink
    English
    arrow-up
    8
    ·
    1 year ago

    skimming through coreutils’ copy.c, emit_verbose is called on line 2627 while copy_reg is called on line 3103 (in the implementation of copy_internal). at least on my machine, touch /tmp/foo && cp -v /tmp/{foo,bar/} returns an error after printing the verbose output.

    • flux@beehaw.org
      link
      fedilink
      arrow-up
      1
      ·
      1 year ago

      Another way to check is to

      strace cp testfile testfile2
      

      and the sequence in which the message is printed and operations performed can be studied.

      It’s perhaps a lot to read, but linux tracing tools are worth learning!