?

Log in

No account? Create an account
FF Sparks (Casual)

Noderunner Stats

So, I ran the stats for Noderunner. Now, these aren't for a full month, obviously, but it's interesting to see. The full RTG information is not all that interesting, alas, because there's not really a true access pattern to snafu, and segfault's access spikes just represent 'waves' of spam mail being received by the mailscrubber.







It's readily apparent that the 15th is when the new antispam solution went live partway through the day on segfault, and you can see that the MX propagated out over the next day, and then I modified the mailservers which segfault scrubs for to toss un-scrubbed mail to segfault for cleaning (so that even the spam-systems which hit backup MXes will get scrubbed). The 4.0 baseline on segfault when I wasn't using it after Firan had moved off to its own machine is just the basic watchdog/monitoring tools, and automated update-checking.

Snafu, obviously, has rather higher stats, being the shell/mail/webserver. Still, I find it interesting that now neither is anywhere near hitting their bandwidth limits, /or/ their CPU limits. Overall, I'm quite happy with the setup at ThePlanet. :)

Comments

Sounds like you've done a big job there. I'm curious, is this just SpamAssassin, or are you also doing something to throttle spammers (via iptables or BSD ipfw)? Throttling may result in some bandwidth improvement.

I haven't found a really good method of throttling yet, but I'm looking into some. Right now, it's several RBLs, some header checks, SpamAssassin with Bayes and DCC/Razor2, Panda Antivirus, all tied together with MailScanner and a few homebrew tools.

That said, considering segfault's eating like 10 gigabytes per month going forward (by my estimate), out of 1000 gigabytes allowed, I'm not /too/ horribly concerned. :)