I've been running SA on my home workstation for almost a year now, and I love it.
Except... occasionally when a lot of mail comes in, my workstation seems to lose interest in what I'm doing, and telnetting in from another machine and running top indicates that spamd is getting most of the CPU cycles. So I'm thinking about offloading at least some of my mail processing to another machine. I've got a 386 based machine which I've build with RH 9 which I called clavin, my workstation is named frodo and also runs RH9. I'm thinking about moving both of these to Fedora core, but that's another topic. The question which is most relevant to this list, is how do I move/copy the SA installation from frodo to clavin and preserve the bayes learning? My second question is a little less relevant but I suspect that this might be a good audience. How should I distribute my mail processing. Here's what I'm running on frodo now: fetchmail runs as a cron job and gets isp mail from my account and my wife's account every 10 minutes. fetchmailrc passes it through spamc. Spamassassin 2.63 is running on Frodo. I'm also running sendmail, which forwards outgoing mail via my ISP, and also processes incoming mail for this domain (denhaven2.homeip.net). I was driven to setting up a separate mail server which I use for mailing list subscriptions and the like after I found that my ISP was overusing blacklists which was preventing mail from certain lists from arriving. Frodo is an smp 686 machine, whereas clavin is a 386 (Cyrix) machine. Is my best bet to just run spamd on clavin, and leave everything else back on frodo? Spamc does support this via the -host arg doesn't it? I think this might be the safest from the perspective of locking my mail spools, or should I look at putting more mail functions on clavin? If so what should I look out for?
