On Saturday 09 April 2005 01:45, Bob Van Zant wrote: > In the worst case you can have a cronjob running that calculates the > size and performs adequate actions for the situation.
This is what we do on our new servers (not the one I pasted times from, yet). We don't believe in cutting off somebody's mail just because they hit some amount of space. If a domain starts eating too much space we send an email summary of disk usage by username to the client and they can either clear up space or pay for more. If their disk usage hits triple what they're paying for, they're sent an email notification, we get a notice sent to us, and their delivery path is then modified to bounce mail, but this is a last resort approach that we try to avoid. > Luckily it takes time to fill up a disk and this "solution" could actually > work for small to medium sized installations. I consider it a perfectly good solution, for any size of installation. The cron job is not heavy, and can be run as frequently as needed. You can also limit the size of each message and monitor for messages coming in at too high of a rate. I personally prefer setting up monitors (i.e. with Nagios) and setting as few bounds as possible, but as you suggest, I do greatly prefer O/S quotas. I also believe that buying more disk space is a lot cheaper than dealing with pissed off customers who've hit quota limits. One other possibility on Linux (and possibly similar solutions exist on other unices) is to use LVM and create a partition for every email domain (or user). I've not tried this, but I hear LVM is dead easy to use. Cheers, -- Casey Allen Shobe | SeattleServer, Inc. [EMAIL PROTECTED] | cell 425-443-4653 http://www.seattleserver.com Please send any responses to [EMAIL PROTECTED]
