When a user tries to go to a website for which the DNS server has not cached an address, the web browser ends up timing out saying it can not find the server. However if the user tries again right away the page loads fine.
What’s happening is that the DNS server is taking too long to resolve the name. While there are several factors that combine to cause this, such as the load on the various root and TLD servers, the only factor you can work around is local performance.
There are two common solutions.
- Increase your available bandwidth.
- Configure your server to forward queries to an upstream DNS server, such as one provided by your ISP or NSP.
For the second solution, if you’re using the Men & Mice Suite, this setting is located in the server’s options window.
There can also be subtle communication problems between your server and other servers. The EDNS0 protocol extension is proving to be problematic on many older (pre 2006) BSD-based operating systems, including both FreeBSD and Mac OS X. There are a few work-arounds:
- You can probably work around this by configuring your server to forward to an upstream DNS server , such as one provided by your ISP or NSP.
- You may be able to fix it by compiling a fresh copy of the BIND name server without EDNS0 support.
- You may be able to disable EDNS0 support in your server’s configuration, or set the EDNS0 maximum packet size to 512 bytes. However, the Men & Mice Suite doesn’t expose this option in the server’s options window, so you’ll have to edit the options manually.