Host keeps closing network connections

[D Waghorn]

I currently have a site server which at completely random intervals shuts all its network connections. It's an A500, HPUX11.0, with standard network card (10/100 port).
The only way, currently, to re-open connections is to remove the rj45 from the server and then plug it back in!
I have checked all the obvious physical and hardware stuff but can find no problem.
I've sniffed the network for collisions, checked the syslog for clues, reconfigured the card (lan0) and run all the available and relevant HPUX networking commands.
I am now at a loss to find the cause of this fault..
Has anybody seen this before?

[T G Manikandan]

I would do a

1.
#lanadmin
check for the collisions on the interface.

lanadmin

lan-->dis-->

2.Make sure that the speed and duplex settings are same on both the ends.i.e server LAN card and the switch.

3.Replace the cable.

4.Make sure there is no duplicate Ip on the network.

Revert with the output of lanadmin.


Thanks
Manikandan

[U.SivaKumar_2]

Hi,

First Have you restarted your switch and observed whether the problem repeats ?

Have you Crimped the Server's ethernet cable on both ends ?

Try changing the cable in if problem persists.

regards,

U.SivaKumar

[Karamanolis Nikos]

Hi,

We had the same on a clients network some weeks ago. Turned out to be a faulty PC net card, which was freaking out an messing up the network...

hope this helps,

[Adam J Markiewicz]

Hi

Mayby it's stupid suggestion, but can't that be ined coredumps?

Good luck

Adam Markiewicz

[Jannik]

I use gpm or lanadmin to check the collitions on the net. To check you duplex setting (i think that is it) use :
lanadmin -sx 0
0 - for the interfase number in you get from lanscan.
To change to say 100 full duplex do the following:
lanadmin -X 100FD 0
and check again after 5-10 sek (have to reset):
lanadmin -sx 0
and check if you still have collitions with lanadmin or gpm.

BR
Jannik

[Ron Kinner]

I presume when you say "...shuts all its network connections." that you mean it stops responding? It is not sending FINs or RSTs is it?

When it fails to respond are we talking WAN traffic or LAN traffic? I assume its all traffic.

Have you changed out the NIC yet? I had to do that the other day for a similar problem.

What sort of switch do you have it on? Is the switch seeing any errors on the port? Could the switch be shutting down the port for some reason? (Too many errors?) Is STP running on the port? Is the switch still sending traffic to the A500. Does the A500 still show up in the ARP table on the switch?

Ron

[Steven E. Protter]

From another Unix system, ping the affected system.

If the ping times have a lot of variance, or they eventually rise in time and do not go down, you have a bad lan card and need to replace it.

It is possible that the card will respond but there are other problems.

run the dmesg command and see if there are any errors related to btlan card.

look at /var/adm/syslog/syslog.log for errors when this happens.

The steps above and the recommendations from prior posts should provide you some clues on whats wrong and how to proceed.

You might want to look into replacing the cable, it might be flakey. You should also report the behavior to network administration, because it could be other physical problems with the network beyond the control of a Systems Administrator.

SEP

[D Waghorn]

Thanks to all:)
(I've tried to assign points, but seem to have problems submitting)
I eventually found the problem, I had just migrated the site to a new address schema, to IP from X25, using a new gateway and switch to the WAN. We use another 'comms box', within this LAN, to perform OSI conversion and act as an async printer server. I had this configured by an engineer remotely during the migration. Unfortunately this is where the gremlin appeared, due to an error. This box was periodically broadcasting for a reply from a remote server which this LAN does not have a gateway to. The broadcast was out of scope and upset the local server.
Thanks to all again:)