[CentOS] CentOS 5 ssloooowww when network unavailable

[Jordi Espasa Clofent]

> It does not seem to be related to ntpd -- or else ntpd is happy with
> the addition of the hostname to /etc/hosts.

Ok. However, the test has been exitous: you know what is NOT 
problematic. This fact reduces the problem.

> Something -- I didn't do it directly -- had already added the hostname
> to the 127.0.0.1 line in /etc/hosts by the time I got around to
> looking there.  I don't know whether it was there before or not, nor
> what I did (boot with the wired network plugged in?  Download updates?
> The timestamp on /etc/hosts changes every time I reboot, so something
> is updating it).  In any case it was not necessary to disable wireless
> or deactivate ntpd to get normal response back.

Ok.

> At the same time, if I do anything that requires DNS (like try to ssh
> to a remote host), all the symptoms return until the DNS query times
> out, at which point all is well again.  So it is DNs's "fault"
> somehow.  (Interestingly, even though some warnings from ntpd about
> unreachable time servers show up in /var/log/messages, they don't seem
> to induce this problem.)

What a curious problem!
You can enable a tcpdump sniffer and put here the output you'll get. 
More useful and easy instructions at: 
http://www.freesoft.org/CIE/Topics/55.htm

-- 
Thanks,
Jordi Espasa Clofent