John Summerfield wrote: > Les Mikesell wrote: >> Frank M. Ramaekers wrote: >>> Sure I understand that, but If I wanted it to be within 15 seconds >>> response, I'd have to check 4times x 60 seconds x 60 minutes x 24 hours >>> per day. Kind of a waste for a file that may or may not be transmitted >>> each day. >> >> Usually any machine that can handle actually doing a job will have no >> problem with the work checking and not doing it the rest of the time... >> If you are concerned about it, a stat() is pretty cheap compared to >> starting a program to do it, so you might write a long-running program >> to monitor for a new file or a timestamp change instead of starting >> one every few seconds. If you really want a file-event driven >> interface you might look at dazuko. http://www.dazuko.de/faq.shtml >> > > > That's not a very good thing to do in a virtual machine (by way of > example). Contemplate doing this in a zSeries box where there are likely > 100 or more virtual machines. If a stat() every 15 seconds causes a problem, even with hundreds of them running at once you need to look for a different machine. Since you are doing it frequently the inode in question will be in cache and this will be a very lightweight operation. -- Les Mikesell lesmikesell at gmail.com