This is the answer to Gateway Down
This is a classic DNS problem. When I looked at /etc/resolv.conf, it looked like this:
domain xyz.com nameserver 192.74.xyz.xyz
Every time a Windows machine tried to access the Unix box, Unix would do a reverse DNS lookup, using the 192.74 address. If the Internet happened to be down, it would hang a long time trying to get an answer - eventually it would give up, but it was too long for their Windows client application, which effectively just gave up too soon. If the Internet was up, the SCO box would get an immediate "no idea" answer, and would continue quickly. I showed the NT folk that if you were patient enough, telnet would actually connect, but the client application they used just wasn't patient.
This isn't a Unix thing: this would happen with NT or just about any OS I know of; they want to do a reverse DNS lookup when a client connects.
My first thought at fixing this was to make the nameserver point at their internal NT DHCP server which, according to the three NT guys, should provide DNS information for the local network as well as knowing how to get to other nameservers when necessary. However, that didn't work. Not my fault: the NT machine knew nothing about its own DHCP clients. The NT folks were surprised, and said they were sure they could fix it, but that it would take a reboot which was agreed by one and all was Not A Good Idea Right Now.
That's OK, I had another way to fix it. First, I wrote a simple script to populate /etc/hosts, and then modified /etc/resolv.conf to look like this:
domain xyz.com # local NT DNS server nameserver 192.168.2.6 hostresorder local bind
That fixed it. We all agreed that the problem should also be fixed at the NT server, but that was their work, not mine.
Got something to add? Send me email.
More Articles by Tony Lawrence © 2011-04-30 Tony Lawrence