I’m suprised this hasn’t appeared on Slashdot yet but I found it interesting. A study of one of the root name servers done by researchers at the San Diego Supercomputer Center shows 98% of queries it handles are unnecessary. 70% of the queries are duplicates, 13% are for bogus TLDs like .elvis, 7% are queries that already have an IP address in them, and 8% are assorted other minor unnecessary crap. If the root name servers were anything like our DNS servers, 90% of the queries would be unauthorized attempts to update zones coming from broken Windows boxes.
I spent the afternoon trying to diagnose some email problems on a client’s LAN. They have a mixed environment of Windows and Unix servers. Their internal DNS server is a Windows 2000 box. Their external DNS server and email server are Red Hat Linux boxes. Sometime yesterday, for no apparent reason they began experiencing a 30-45 second connection delay between the email server and their email clients (mostly MS Outlook – yes, I’ve warned them that Outlook is the single biggest propagator of viruses in the known Universe but they still use it).
We checked all the usual suspects. No DNS problems, no recent changes to BIND or sendmail.cf, no sign of DoS attacks, rootkits, viruses, etc. I’ve seen a similar problem caused when the Timeout.ident value in Sendmail is set too high but this was not the problem either. For now, it’s a mystery but further experiments will hopefully turn up the cause.
Another week already gone! After solving our Verio DNS problems earlier this week, I spent the rest my time working with PostgreSQL. It took a couple of patches but I got it compiled and running on our Sun Ultra 10 with Red Hat Linux 6.1. I added a new 13gig drive just for the database. That should be enough space to get started. I’ve got our new web server (an Intel/RH61 box) set up with Perl DBI and PHP interfaces to the database server. Meanwhile, I’ve been working with Erin on our first web database application. I sneaked out of the office for a few hours today to catch the Georgia O’keeffe exhibit at the Dallas Museum of Art. It’s over in a couple of days and I’d promised Susan we’d go see it. It’s definitely worth seeing (but skip the audio guide!).
The problem of the day is in-addr.arpa delegation. We’ve been migrating some of our older servers to IP addresses in a new C block that was recently assigned to us by Verio. We’ve been having all sorts of unexplained lookup problems with those servers and finally discovered the cause today. Verio never properly delegated the authority to our DNS server for the new C block. It took a full day of debugging and playing with nslookup and, of course, turned out to be something obvious and simple (doesn’t it always?). Verio submitted the delegation update to Arin tonight. Now I’m a full day behind on my schedule for this week but at least I can get back to more interesting work now.