« Home | In the midst of a vast wilderness, far off, standi... » | Fun Toy Banned Because Of Three Stupid Dead Kids. » | First there was the working artificial eye, and no... » | I used to joke about how my main non-free ISP was ... » | They call it PMS because Mad Cow Disease was alrea... » | Excellent, the next Survivor type show deals with ... » | Oddities. I noticed today that my top line of li... » | Ahh, memories of Computer Science class, junior ye... » | I'll be damned. The site looks almost identical i... » | When I first read the headline for this article ab... »

I worked 17.5 hours yesterday. Took a half an hour break for dinner between jobs, and that was about it for rest. In Ann Arbor, most of the client machines could not connect to anything but the file server. Knowing that the file server uses Appletalk as the protocol and TCP/IP is used for everything else, we narrowed down the list of possible problems. Checking out the error logs on our servers showed stuff like this over and over and over:
Aug 16 18:54:45 beer.pub.umich.edu dhcpd: DHCPDISCOVER from 00:50:e4:30:6f:24 via eth1
Aug 16 18:54:45 beer.pub.umich.edu dhcpd: DHCPOFFER on 141.211.148.76 to 00:50:e4:30:6f:24 via eth1
Aug 16 18:54:45 beer.pub.umich.edu dhcpd: DHCPREQUEST for 10.0.1.5 from 00:50:e4:30:6f:24 via eth1
Aug 16 18:54:45 beer.pub.umich.edu dhcpd: DHCPNAK on 10.0.1.5 to 00:50:e4:30:6f:24 via eth1

In basic terms, this is what was happening: a system comes online and tries to DISCOVER the DHCP server, the server OFFERs a good IP from it's pool, the client REQUESTed a bad IP, and the server does Not AcKnowledge that request. Then the cycle repeats. This made no sense to all three of us working on the problem. We tried restarting our DHCP server, we tried using a different server, etc etc. FINALLY, Sat realized that someone had an Apple AirPort they wanted to hook up. Searching the Apple Tech Info Library, we found out that the AirPort acts as a DHCP server! It was accepting all of the requests, and just giving out bad information! Took that off the network, and everything returned to normal. And it only took 8 hours to figure that out.