Name resolution takes too long - lots of time outs

After upgrading to suse 9.3 from 9.1 the name resolution is spotty.
When connecting to remote hosts on the network, the name resolution
takes too long and times out.
Eg., when I do "telnet mail.yahoo.com 80", the name resolution takes
too long. But if I try the same command after 3-4 minutes, it works.
But becomes unreachable after a few minutes and then starts working
again.
It is as though the name resolution is taking a long time to resolve
the name, which it eventually does seem to. But is not fast enough for
browsers and they eventually time out.
traceroute does work, but is not able to print hostnames along the
hops.
nsswitch.conf, resolver.conf all in order.

Any help appreciated.

On 12 Jun 2005 07:58:35 -0700, hemanir@netzero.com wrote:
> After upgrading to suse 9.3 from 9.1 the name resolution is spotty.
> When connecting to remote hosts on the network, the name resolution
> takes too long and times out.
> Eg., when I do "telnet mail.yahoo.com 80", the name resolution takes
> too long. But if I try the same command after 3-4 minutes, it works.
> But becomes unreachable after a few minutes and then starts working
> again.
> It is as though the name resolution is taking a long time to resolve
> the name, which it eventually does seem to. But is not fast enough for
> browsers and they eventually time out.
> traceroute does work, but is not able to print hostnames along the
> hops.
> nsswitch.conf, resolver.conf all in order.

Show us the output from
cat /etc/resolv.conf
grep host /etc/nsswitch.conf
cat /etc/host.conf
head /etc/hosts
route -n

PS:
Please read http://www.catb.org/~esr/faqs/smart-questions.html
While waiting for an answer.

The section I have in mind is
Describe the problem's symptoms, not your guesses

On Sun, 12 Jun 2005 07:58:35 -0700, hemanir wrote:

> After upgrading to suse 9.3 from 9.1 the name resolution is spotty.
> When connecting to remote hosts on the network, the name resolution
> takes too long and times out.
> Eg., when I do "telnet mail.yahoo.com 80", the name resolution takes
> too long. But if I try the same command after 3-4 minutes, it works.
> But becomes unreachable after a few minutes and then starts working
> again.
> It is as though the name resolution is taking a long time to resolve
> the name, which it eventually does seem to. But is not fast enough for
> browsers and they eventually time out.
> traceroute does work, but is not able to print hostnames along the
> hops.
> nsswitch.conf, resolver.conf all in order.
>
> Any help appreciated.

There seems to be a fairly well documented problem with certain dns
servers timing out a couple of times before servicing requests from
certain linux boxes. After installing Mandrake 10.1 on public access
internet computers at the local library, I was seeing times around 45
seconds. I stopped IPv6, and times went down to 12 seconds. Added 'options
timeout:1' to /etc/resolv.conf and times became 1 or two seconds.

In the Usenet newsgroup comp.os.linux.networking, in article
<1118588315.202499.225070@o13g2000cwo.googlegroups .com>,
hemanir@netzero.com wrote:

>After upgrading to suse 9.3 from 9.1 the name resolution is spotty.
>When connecting to remote hosts on the network, the name resolution
>takes too long and times out.

Did you disable IPv6? Not in the fscking browser, but in the kernel
itself.

alias net-pf-10 off

in /etc/modules.conf. SuSE wants to be cutting edge, even if the rest
of the world isn't ready for it. Do a groups.google.com search paying
attention to the alt.os.linux.suse newsgroup, and you'll see lot's more
about this problem.

>Eg., when I do "telnet mail.yahoo.com 80", the name resolution takes
>too long. But if I try the same command after 3-4 minutes, it works.
>But becomes unreachable after a few minutes and then starts working
>again.

That's a name caching problem. Are you also running the nscd (Name
Service Caching Daemon)?. The correct solution is to disable IPv6,
so that you don't have the problem in the first place. What's happening
is that the name query is taking longer than needed thanks to the SuSE
configuration, but the information eventually reaches the cache. When
you look again, it's cached, and is available quicker.

>traceroute does work, but is not able to print hostnames along the
>hops.

traceroute, ping, and a lot of other network applications all have a -n
option for a reason.

Old guy

Disabled ipv6, turned off the name cache server daemon - nscd - off.
Have not tried the timeout option in resolv.conf.
It definitely is a name resolution/cache problem.
Updated bind libs from suse updates, problem persists. Have not done a
fullscale investigation with tcpdump or ethtoool. Looked at a few
outputs from tcpdump, but did not have the bandwidth to continue.
Applying a few more patches from suse.