Silly question - how does the "mv" command work?

Shell Programming and Scripting
1

I know this may sound really elemental, but I'm trying to figure out if I'm correct.

I have a script that moves a file from a temp directory to (what I am calling) a pickup directory.

On another machine, I have this "other program" that scans the contents of the pickup directory for a specific file name. If the file name exists, the program goes forth and does a bunch of routines ... (really not plausible to my question)

However, I have a feeling that this "other program" is grabbing this file too soon as the "mv" hasn't successfully completed writing the files contents; perhaps just the filename.

I did a test on the "other program" and the scan looks for the said file (in a while loop) approx. 22,000 within 10 seconds (a timeout).

The size of the "moved" file varies but is averages at about 200-300 bytes.

Can someone tell me if the move command first creates the file then writes?

Thank you!

2

If both directories are in the same file system only the inode table is updated, the file does not actually move

---------- Post updated at 12:01 PM ---------- Previous update was at 11:54 AM ----------

If the two directories involved in the mv are on different file systems, the a copy and delete is done
However, unix systems will show the file in a ls as soon as it is opened for writing, so your second program should confirm that the file is not in use (use fuser), otherwise you may process a partial file.

3

Actually, the file is being moved to an NFS mounted directory. Any ideas?

4

When you mv the file from its original location to the pickup location, when finished, create or move a second file with the same name as the data file, but a different extension (eg data.dat and data.job)
Have your second process only process .dat files for which there is a equivalent .job file.

5

Can the scanning program be configured to skip files beginningin . ? That way, you could upload it as ".filename", so it will be skipped, then rename it to "filename" once ready to be processed.

6

However the inode does not change when moving a file within the same filesystem:

$ ls -i file1
    139787 file1
$ mv file1 file2
$ ls -i file2
    139787 file2

Is the same as creating a hard link, and deleting the original reference:

$ ls -i file1
    139787 file1
$ ln file1 file2
$ rm file1
$ ls -i file2
    139787 file2