We have a third party tool in UNIX to kick off a 'file copy' job based on a file existance. If a specific file exists in an UNIX directory, another process should start copy the file into another system for further processing. The issue is, the copy job is starting as soon as the file exists in the directory but the file creation process is still writing records into the file.

How to identify the file is closed before starting the 'file copy' job.

We have a UNIX script to check the file size for every 30 seconds and compare the previous and current size, if no difference in size
we assume the file is closed. This approach does not work if the file creation process is delayed or slow due to network etc..

Any help is appriciated.

Thanks
K. Lakshmanan

12345

If no lsof is on your system, then you may have fuser instead. Does pretty much the same.

I admit the following is not a good logic , but i'm not able to control sharing with you.

Place some type of identifier (special string) after first process finished writing into the file. and force the second process to look (grep) for that identifier (special string ) and continue the processing ....... copy ............whatever ...

My favorite way to handle this is to arrange for the producing program to produce two files: the real file, followed by a tiny flag file. When the flag file appears, you know that the first file is ready.

A way I use to find growing log files on an unfamiliar system is:

touch /tmp/test
find /path/to/suspect -newer /tmp/test

COUNT=`fuser /path/to/file | wc -c`
if [ $COUNT -gt 0 ]
   then echo file is open
fi

We also like to use flag files, like Perderabo said. Just have your copy script #1 copy your data file into the directory and then upon success, touch a flag file. Then have copy script #2 check for the existence of the flag file. If the flag file is there, copy the script. If not, wait n minutes.

You also want to remove the flag file, either at the beginning of script #1 or the end of script #2.