Red Hat Bugzilla – Bug 76543
name to IP resolution issues
Last modified: 2016-11-24 10:15:17 EST
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.1) Gecko/20020830
Description of problem:
Resolver. General problem with name to IP resolution.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
See Additional information section for configuration information
Actual Results: See Additional information section for configuration information
Expected Results: names should be resolved to IP via files when files specified
first. BIND should resolve locally first when local host names are used.
There appears to be a general issue with name to IP resolution in RedHat Linux
8.0. This is present in a number of different situations.
Senerio 1. Initially I was working with a simple files based network setup.
That is my system names were listed in /etc/hosts both as fqd names (dot 3) and
short (just server name) name alias. Both /etc/host.conf and /etc/nsswitch.conf
were set for files first and then dns. My IP's nameservers where set in
/etc/resolv.conf. The first thing I noticed is that when trying to connect to
systems on my local network with this setup ssh and telnet attempt to resolve
via DNS first. Only after the attempt times out or otherwise fails will it fall
back to files. This is true both with short and fqd system names. Oddly enough
ping is not effected. With ping the hostnames appeared to immediately resolve
directly from /etc/hosts without an internet DNS lookup attempt. I am throughly
convinced the behavior under this senerio is a bug.
Senerio 2. Decided to setup a DNS (BIND) server for my local network. Slightly
different resolution pattern observed. All fqd (dot 3) local network system
names resolved without issue for ssh, telent, and ping. However, short names
did not appear to resolve properly. Encountered delays before they would
resolve via local DNS (I had commented out entries in /etc/hosts and had
internet DNS disabled to ensure I was resolving only through my DNS). At first
I thought this was a setup problem with my DNS server but I found that when I
tried ping the system names resolved immediately without delay even with just
short (host) names. It seems almost as if the domain name is not being
appropriately taken into account when using ssh and telnet during the hostname
resolution. I tried entering the domain into /etc/resolve.conf but there was no
change. Either way ssh and telnet attempted to contact outside DNS servers
(this could be evidenced in firewall logs after setting the fw to drop outbound
DNS at the gateway) before using the local zone information, however, ping
immediately resolved using the local zone information without attempting to
contact internet servers.
Sernerio 3. The final thing to report is that I have just observed that mail
(sent from anacron to root on the local system in this case) appears to be
subject to external DNS resolution attempts when using the files based network
setup described in senerio 1.
Not component is not properly classified but I could not determine proper
library. Please reclassify.
This bug also appears as bug 77538
RedHat: (can we get || is there) a master bug for this issue ?
I can confirm this bug. In fact this bug has become a show
stopper for our RH7.3 to RH8.0 upgrades. This problem does
not show up in RH7.3, FYI.
Please consider changing the Priority of this bug from normal
FYI: I think this bug is better worded and addresses the
more general problem than bug 77538.
This general bug is related to /etc/nsswitch.conf. The command:
rpm -q -f /etc/nsswitch.conf
shows that /etc/nsswitch.conf is part of the glibc component.
And the client/user-side name resolver is found in libc/glibc.
In addition to changing the Priority to high, I recommend that
the Component be changed to 'glibc'. I believe that normally
the Bug author (email@example.com) is able to make these
changes. So firstname.lastname@example.org, please make those changes
in the 'Edit Bug Information' section and click submit below.
NIS/ypbind is suspected to not work correctly as a client in RH 8.0. I have
received reports from co-workers and will provide additional information during
week of 6-January-2003. Having read Bug# 77539, and this bug I believe there is
a serious defect in a name resolution library.
correction to my previous comment: it should have referened but# 77538
Updating component and severity information.
I think bug 77538#c9 and bug 77538#c10 may help in diagnosing the problem.
There is an ugly hack-a-round for this bug:
Duplicating each ip-host entry in /etc/hosts with an 2nd
trailing dot entry allows one to hack-a-round this problem.
This is a gross hack. This glibc bug really does need to
be fixed. But while we wait for this bug to be fixed,
you can try this hack.
See also bug 84105:
excessive IPv6 lookups done even when /etc/hosts resolves to IPv4
I appears that this name to IP resolution problem has been resolved in
A clarification on:
You should NOT install those RPMs on a production system. Rawhide
is raw bits. Those RPMs were only in relationship to various DNS
issues. Those rpms have a number of non-DNS related problems.
For example, they cause the rpm command to dump core. They did,
however, resolve the DNS issues, with the possible exception of
excessive IPv6 lookups.
There are a few misconceptions in the original report. The order specification
in /etc/host.conf is pointless nowadays, this is what we have NSS for. The file
is read but order is not used.
Regarding the stall and DNS lookups, no recent release should have any problems
*** Bug 81737 has been marked as a duplicate of this bug. ***