[SUMMARY] Q: intermitant name resolution problem

From: Tuna Tugcu (tuna@tunafish.cmpe.boun.edu.tr)
Date: Wed Apr 22 1998 - 06:47:44 CDT

The original question:
> I have just installed 2.5.1 on a sparc. I've set in nsswitch.conf hosts
> entry to dns and files. I've also set domain and nameserver in
> resolv.conf. I did multiple successful ftp and telnet connections. Then,
> after some hours, I started getting failures in name resolutions. I get
> correct responses thru nslookup, but telnet and ping fail. I noticed that,
> though it doesn't make sense, when I touch resolv.conf, I can make exactly
> one successful name resolution; then it starts failing again. It doesn't
> always do a successful lookup just after a touch, but generally, general
> enough to make me state it here, this is true.

   First, I tried writing resolv.conf from scratch as Ruiz Carrete
proposed, but that was not the problem. I tried nscd stop, it didn't
solve the problem either. But killing the nscd daemon solved it. So
I just commented out nscd in the rc scripts.

   Tuna Tugcu

Special thanks go to:
Igor Schein
Joel Lee
Brion Leary
Tom Erickson
Jim Robertori
Josť Gpe. Ruiz Carrete
James O'Reilly SSDS, Inc.

Here are the answers:
From: Igor Schein <igor@bob.air-boston.com>

Where is your DNS server? Is it also a Sun machine? I have the same
problem here, because primary DNS server is an NT machine, and I haven't
set up a secondary DNS server. So check if you have a secondary DNS
listed in your /etc/resolv.conf
From: Joel Lee <jlee@thomas.com>

Try to turn of the nscd process and see if that helped. If that helped,
try commented out the line with "enable-cache host ..." and
restart nscd.
From: "Brion Leary" <brion@dia.state.ma.us>

It may be a problem with nscd(1m). It is a Name Service Cache
Daemon. try retarting it -
        /etc/init.d/nscd stop
        /etc/init.d/nscd start

On our firewall machine I have disabbled it - I do not run it at all.

I have seen it fail to resolve addresses after a change on a DNS
From: Tom Erickson <Thomas.M.Erickson.1@gsfc.nasa.gov>

I suggest you stop and restart the nscd process. This process
caches naming services (more than just nslookups).
From: Jim Robertori <jimr@lucent.com>

Sounds like you have trouble upstream with a DNS server. When you touch the
file, you are actually updating the time stamp. This indicates to DNS that it
should check with an *AUTHORITATIVE* server which it does that one time. After
that, it is probably checking the non-authoritative servers. What's in your
resolv.conf and are they still correct.
From: "Josť Gpe. Ruiz Carrete" <jruiz@dali.itdgo.mx>

Some time ago, I had a problem a bit similar than yours. I was testing the
name configuration server
because this machine was the one which can not resolv names for itself, but it
can resolv names for others ones. I checked the resolv.conf and nsswitch.conf
and they semmed to be ok. Finally, I rewrote
the file resolv.conf and..... I can believe it worked, yes its real. This was
a magic solution but was like.

What I think is the file resolv.conf had some strange non-printable characters
into it. You could try this, is fast and you have nothing to loose....
From: "James O'Reilly SSDS, Inc." <jpo@tesla.wcpss.net>

1. Check sunsolve for any named patches.
2. Try using the debug tool: nslookup -d2 server_name
3. Upgrade to BIND-4.9.Latest_Version which replaces suns' (named,nslookup)

This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:12:38 CDT