About 9 days ago my main server was rebooted as part of a UPS test where it is hosted. What I didn't notice until today is that the fetchmail process that normally runs on that box to grab my jzawodn@linux-mag.com and editors@linux-mag.com mail had not been started.
I should have noticed this on day #1 or maybe day #2 if I was a bit distracted. But nooooo. It took NINE F@%*ING DAYS for me to realize "hmm, I haven't seen much Linux Magazine mail in a while." In fact, when I went to check, I realized that I hadn't seen ANY of it for quite a while.
After a few seconds of head-scratching, I realized what was going on. I switched sessions, started up fetchmail, and started a tail -f .fetchmail.log process. Here's what I saw:
1548 messages for jzawodn at mail1.via.net (10910467 octets).
Ugh.
Luckily, 95% of it was spam and SpamAssassin handled it just fine. But that still left a surprising amount of mail to plow thru, including several notes from the Editor in Chief. It's no wonder he picked up the phone and called me last week to find our where my columns were. Not only was I not responding to his e-mail, I wasn't even reading it.
Ah, well. This shouldn't happen again.
Posted by jzawodn at February 19, 2004 04:21 PM
I think I had my fetchmail startup and shutdown in the ip-up and ip-down scripts.
Classic Unix mistake — start a daemon, check it works, then happily forget to put start/stop commands in appropriate scripts.
Cron started supporting an @boot time at some point which I find invaluable for this sort of thing. I have a ton of utility scripts which I leave running to do things like monitor log files, advertise Unix services to rendezvous clients, etc. which I was always forgetting to restart since the systems were rebooted very infrequently for kernel updates.
Classic Unix mistake [...]
I make this mistake far too often. It's a good one to remember when installing new processes, though.