Jason,
Thanks for the brilliant suggestion. With a little research I whipped up a Perl script for the read end of the pipe. It writes the data locally, then sends it down a socket to another script on the remote machine for replication. I played around with rsh and ssh, but they were lagging pretty badly (5-10 seconds to return) on a regular basis (roughly 1 out of every 20 instances). Not only is the socket really fast and reliable, it also prevents the need to spawn another process per write. Time will tell, but I believe this is the right solution.
I appreciate everyone else's suggestions as well.
MR> Please provide me with tips on how to replicate a single file to a MR> separate machine as changes are made to it.
If your application truly only supports logging to a file, then consider seeing if it will continue to log when you replace that file with a named pipe. Connect the read end of that pipe to a program that does whatever you want; you could even connect it to an instance of ssh or nc to a remote machine.
- J<
Sincerely,
Matthew Roth InterMedia Marketing Solutions Software Engineer and Systems Developer