OK, here's what I'm trying to do: I am trying to write a script that will be given a command, and that script is to run that command on a number of other remote machines in parallel, then collect the STDOUTs from these remote executions and report back to the user running my script.
Let's assume the remote executions themselves are being done with ssh. If I didn't want to run the various remote executions in parallel, I could simply use backticks to execute ssh with appropriate arguments plus the command, then wait for the backticks to complete. But I need multiple ssh's to be run in parallel. My first try was to perform each ssh execution in a separate thread. I then used open() with a pipe to execute ssh and read the remote execution's STDOUT. Then it slowly dawned on me, open() with a pipe symbol causes a fork() behind the scenes. That would mean I have overhead of first starting a thread then the thread causing a fork(). That doesn't sound good. I thought about forking, then having the fork exec() ssh, so the full Perl interpreter in the fork would be replaced by just the executing ssh process. However, I don't think I could get the STDOUT of the executing process back. Should I spawn a thread from each remote execution, then have the thread use backticks to execute ssh? My head is beginning to hurt. Does anyone have any suggestions? By the way, I know about how to present the arguments to system()/exec() so that Perl doesn't have to spawn a shell to execute the command. Merrill _______________________________________________ Perl-Unix-Users mailing list [EMAIL PROTECTED] To unsubscribe: http://listserv.ActiveState.com/mailman/mysubs
