So so weird routing problem

[Calin Maniu]

I have a HPUX 11.11 server (rp5470). When it boots up it is accessible on our internal WAN - it's local LAN is a class C range. However after ten minutes it drops off the WAN. It is still accesible on the Class C LAN but doesn't want to route.

The routing table stays the same all the time and I am having trouble spotting any process that may be doing anything. The network configuration must be good because it is available for a small time on the WAN and is always available on the LAN!

Any ideas??

Thanks,

Sam

[steven Burgess_2]

Hi

Check

netfmt -t 30 /var/adm/NETTL00.log > /tmp.nettl.out

Are there any recent errors ?

Once the HP box disappears off the WAN have a look at your gateway router. Can you see the mac address and ip of the HP server in it's arp table

sh ip arp

Ensure that these match the ip and mac of the HP box

lanscan

To get the mac address of your HP box

It does look as though you may have a duplicate address somewhere

HTH

Steve

[Paula J Frazer-Campbell]

Hi Sam

How many NIC cards ?

Any errors in syslog.log

Any changes made to routing eith on the server of routers?

Sounds like either a NIC card or routing problem.


Paula

[Calin Maniu]

There are two cards but only one in use. There are no errors in the syslog being reported. Routing changes are not happening on either the router or the server itself.

I will try the netfmt and check arp results on the gateway now.

Thanks!

[Paula J Frazer-Campbell]

Sam

Further questions
:-

Is this a new build or is it occuring on an otherwise well behaved machine?

If you have a spare NIC card configure it and use that so that the existing NIC is eliminated from the equation.


Paula

[Mark Greene_1]

It could be that you have a duplicated IP address on the WAN. Ping the IP address on different sub-nets and see how much the response time varies. If you have any sort of sniffer software or equipment, try that as well.

HTH
mark

[Calin Maniu]

The netfmt showed up clear and the arp on any machine on the HP boxes LAN reports the correct interface. A traceroute on the local server shows the server itself and then it looks as if it thinks it has completed!!

[steven Burgess_2]

Hi

Do you have access to a server that doesn't reside on the WAN ? Try a traceroute from there. There is an option with ping that will show you each station that the packet passes through.

man ping

sorry, can't remember the option

Try this from a device on the WAN

Also , take the server off the network then attempt to ping the ip on the LAN and the WAN

You say the server isn't pingable on the WAN, can you get to the another gateay router on the WAN and do

sh ip route

HTH

Steve

[Calin Maniu]

I have access to another system on the same LAN and using the same gateway it can reach the WAN. I have noticed something odd - when I traceroute from this odd HP server I get an "!N" appearing after the timeout indicators and it returns a command prompt immediately. A traceroute from the other server ges through.

Does anyone know what the "!N" means?

[Mark Greene_1]

Much to my surprise, neither of my 11.0 servers have a man page for traceroute. I lifted this from the aix man page:

"The traceroute command prints an ! (exclamation mark) after the round-trip time if the Max_ttl value is one hop or less. A maximum time-to-live value of one hop or less generally indicates an incompatibility in the way ICMP replies are handled by different network software. The incompatibility can usually be resolved by doubling the last Max_ttl value used and trying again. "

Can you trace to the server in question from a server on a different sub-net using a different gateway?

HTH
mark

[Calin Maniu]

Turns out the !N just means destination netowrk unreachable.

However. I remoed the default entry in the routing table and readded exactly the same and miraculously everything started working. But then it stopped after only a few minutes.

The netfmt log showed that the server keeps seeing it's NFS resources as being unreachable every few minutes as well.

[steven Burgess_2]

Sam

Are your routing table entries similar to the same server on the LAN that is having no problems ?

Steve

[Calin Maniu]

The other machine I have access to is a PC, but yeah - same default gateway, with no other routes configured.

[Duncan Edmonstone · Accepted solution]

Sam, what is the gateway router? Its not a firewall is it? If so, you may be suffering from dead gateway detection...

Have a look at the 'real' routing entries using ndd:

ndd -get /dev/ip ip_ire_status

make sure your gateway isn't marked as DEAD.

See the first part of my post here:

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xe26146ff9277d511abcd0090277a778c,00.html

HTH

Duncan

[Calin Maniu]

Yep - that gateway is marked as Dead. we are running HPUX 11.00 as well so I assume I need to hunt for an updated 100Base-T driver.

Thanks for your help!

[steven Burgess_2]

sam

Without access to a router or another machine on another subnet I can't really help you much more as we can't see how the router see's a route to the machine. My PC knowledge is extremely minimal. Do you have a network administrator who will be able to help

HTH

Steve

[steven Burgess_2]

Duncan

'wicked' well done mate
I'll add that one to the archives

Steve