Shell Programming and Scripting

// Redhat

I am running a cron with cp every 15 minutes.

Code:

cp /p2/arch/log/* /p2/bkp

What I need is to not to copy the file again if the filename already exists under /p2/bkp

The reason is that the file size under /p2/arch/log/ will be reduced (contents truncated), so I need to keep the files copied over for the very 1st time.

In that case, I keep the original files w/o losing the contents.

Please advise.

Hi.

You can use the -n or --no-clobber options.

Note, however, that if a file is being written into /p2/arch/log/file when the cron job starts running, you might copy the initial contents of file to /p2/bkp/file and never pick up data written to the end of the file after the copy completes.

Are files in /p2/arch/log being updated after all of the data that is initially written to them has been completed? If not and /p2/arch/log and /p2/bkp are on the same filesystem, did you consider just making a hard link from /p2/arch/log/file to /p2/bpk/file?

Pardon, a link as a backup??
If there is a risk that a file is being written during the backup then I would rather ensure the files are a minute old

Code:

find /p2/arch/log -type f -mmin +0 -name '*' -exec cp -n {} /p2/bkp \;

Sorry... I should have been much more verbose about the environment needed to use a hard link as a backup method.

I once worked in a group where a logging daemon was started when the system booted and ran forever to log errors that would occur. It opened the log files with something like:

Code:

        fd = open(logname, O_CREAT | O_WRONLY, S_IRUSR | S_IRGRP | S_IROTH);

(note that it is opening a file for writing with no write permissions on the file it creates).

If the open() failed with an EPERM error, it looked to see if another daemon is already running (and quietly exited if it was). Otherwise, it incremented that day's log sequence number and tried again.

At midnight, it closed the current log file, opened a new log file for the new day, hard linked any log files created the previous day to a backup directory, and removed any log files in the log directory that were more than a week old (leaving the link in the backup directory untouched).

Another process ran every few days archiving the log files in the backup directory to tape and removing log files more than three months old from the backup directory.

This scheme worked well for that project, but it assumes a lot about how the logs are created and used that is atypical for most loggers today.

Code:

cp -n or cp -no

Flag doesn't work.
Which Flag should I use?

Assuming that you just want to copy files from the directory /p2/arch/log (and not from subdirectories of that directory), you could try something like the following. It is untested, but should come close to what you were trying to do:

Code:

cd /p2/arch/log
for f in *
do	[ -f "$f" ] || continue		# If this isn't a regular file, skip it.
	[ -e "/p2/bkp/$f" ] && continue	# If a backup already exists, skip it.
	cp "$f" /p2/bkp			# Make a backup copy.
done

This should work with any shell based on Bourne shell syntax (e.g., ash, bash, dash, ksh, sh,zsh, etc.), but will not work with csh and its derivatives.