I can only emphasize what Don Cragun has already said. You didn't post your script therefore, instead of offering detailed advice, i will only make some general remarks. You are invited to post your complete script if you want to get better and more specific help.


I think we have here a (common) misunderstanding about what "export" does, so let us address that first.

Every process has its own set of variables (and their respective values), which form the so-called "process environment". Processes form hierarchies, where one process (the "child process") is called by the other (the "parent process") and it would be good to have a mechanism which passes the process environment of the parent on to the child.

Exactly this does the "export" keyword. Issuing an "export <variablename>" flags the variable to be inherited by an eventual child process if one is invoked. As long as you stay inside the process there is no difference. You also do not need to "export" the variable again if its value changes. A variable is either "marked for being inherited by the child process" or it is not, but the inheritance takes place at the time the child process starts and with the value the exported variable has by then.


No. If your script is called twice these two will be (maybe simultaneously running) instances, which will - from an OS point of view - not interfere with each other. They will have two separate environments and will run alongside each other.

Of course, just because they are indenpendent in the eyes of the operating system that doesn't mean they couldn't interfere at all. If they use the same resources - they probably issue commands against the same database, which will cause the DB to change its state - they can cause problems for each other. But this has nothing to do with their process environments, but rather with a competition for common resources, where the "common resource" in this case is the database or some vital part of it.

My suggestion is to make sure the script is executed only once at a time. Usually you do this by creating a PID-file at the begin of the script and removing it at the end. See the following sketch:

Code:

#! /bin/sh

typeset fRun="/var/run/myscript.pid"

if [ -e "$fRun" ] ; then
     echo "ERROR: can only be run once" >&2
     exit 2
fi
echo "$$" > "$fRun"

# ... rest of your script goes here ...

rm -f "$fRun"
exit 0

Depending on your shell you may put the removal of the PID file into a service routine for "trap 0", which will enhance the runtime security, but not change the basic, underlying logic.

I hope this helps.

bakunin