Flashes of Panic

There were more transient email failures this morning. Like last time, parts of the system worked, other parts didn’t. Restarting the daemons didn’t help. Watching the log file (tail -f /var/log/maillog produces a nice stream of log messages as they are appended to the log file) showed that there was a lot of stuff going on successfully (it is infuriating to see the amount of time that server spends beating off brazen spammers—it’s like watching someone trying to do patient embroidery in a mosquito swamp.) And yet stuff was timing out. (I blame the spammers. I always blame the spammers.)

A theory from a helpful colleague suggested DNS timeouts. (That is, the service which resolves internet names like mail.flashesofpanic.com into numeric IP addresses was responding so slowly that my mail servers were giving up.) Our service provider does that for us; maybe they’re getting swamped. Maybe I need to be doing that down here? Another service to administer when I don’t quite understand the subtleties of the configuration files (and yes, I’ve got the cricket book.)

One thing that does have to happen is making Raven our backup mail exchange (MX) instead of our ISP’s mail server. The ISP guarding the back door is forced by their situation as a service provider to be more lenient about suspected spam, and a lot of trash is relayed in by that route. If I have responsibility for both primary and secondary MX, I can be equally strict on both of them, and hopefully lighten the load on both.