System draging because of DNS issue

[Marty Metras]

I have been looking for a cause for the System being so slow. Being that CPU, I/O and Memory ussage were next to null. It did not make sence until network guy called me and told me the Primary DNS server was offline. Ahhhh!
Well it seems that this HP-9000 with HP-UX 11.0 takes for ever to switch over to the secondary DNS server. The network guy knows little about HP-UP. So I know you guys know a lot.
My question is where do I look to verify that both DNS server are beeing seen and is there a place to set the timing on the HP for how long it looks for the the DNS server before it looks for the other one?
So where should I start looking.
Marty

[James A. Donovan]

You can try modifying the hosts line in your nsswitch.conf file.

e.g.
hosts: dns [NOTFOUND=continue UNAVAIL=continue] files

See the nsswtich.conf man page for more details

[SS_6]

First look at the following files try to ping DNS servers entries from resolv.conf file
/etc/resolv.conf and /etc/nsswitch.conf

[Jeff Schussele]

Hi Marty,

Most systems have multiple DNS servers defined in the /etc/resolv.conf file.
A quick easy fix would be to flip the order & make the misbehaving one the last one. They are searched from the top down. Change is immediate & nothing else need be done.

HTH,
Jeff

[Steven E. Protter]

The command on windows is nslookup

The possible commands on hp-ux are:

nslookup
dig
This will with command options get you started.

Took a while to get this post.

SEP

[A. Clay Stephenson]

There are two fields (retry and retrans) that you can add to /etc/resolv.conf. Man 4 resolver for details.

However, my favorite method (which is a holdover from pre 11.x HP-UX which did not implement the above values -- the timeout was 30 seconds, off the top of my head) is to run a cron job every few minutes that does an nslookup for each of the nameservers listed in resolv.conf. If an unresponsive nameserver is found, move it to the end of the nameservers list in the file and the fastest responder becomes the first.

[Marty Metras]

The DNS server is back up so testing will wait for now.
Here is what I changes.
/etc/nsswitch.conf
# 04/14/04 MEM Changed because hangs if Primary DNS fails
# hosts: files [NOTFOUND=continue] dns
hosts: dns [NOTFOUND=continue UNAVAIL=continue] files

/etc/initd.conf
# 04/14/04 MEM added -TCP_DELAY
telnet stream tcp nowait root /usr/lbin/telnetd telnetd -b /etc/issue -TCP_DELAY

ping, nslookup, and dig all returned what they should have. Or they do now. The changed needed to be done.
If it acts up again I will see if Bob is here first. I checked the log from last week and Bob was here then too. Never found the problem then.
/etc/resolv.conf only has the 2 name servers IPs.

/etc/hosts only has a few of my pet servers and the printers.

New question. If I restart the network services using "inetd -c', will it break the connections to the users?
Marty

[A. Clay Stephenson · Accepted solution]

No, established connections are not harmed by inetd -c which really sends a kill -1 to inetd.

[James A. Donovan]

"inetd -c" simply causes the inetd process to re-read the inetd.conf file. It will not disconnect anyone currently connected.

Side note: I found the retry and retrans parameters for the resolv.conf file documented in the 3N section (man 3N resolver). They look very useful, I wonder why they aren't mentioned in section 4.

[Bill Thorsteinson]

To see if a particular server is resolving
requests provide the server, by ip or name
as the second parameter to an nslookup request.

nslookup somehost dnserver

This will force that server to be used. If it isn't responding you will get an appropriate message.

Quick fix is to remove the server from the
/etc/resolv.conf file.

You might consider configuring software like
dnsmasq to provide a local DNS proxy.