TCP/IP stops working on SCO Release = 3.2v5.0.4 1

[wpdavids]

My SCO server will run fine for about 1 week then all users will be unable to connect with telnet. The server has about 40 users that all experience a frozen telnet screen. I try to ping the server from a PC or other Unix machine, but I get no replies.
Finally I go the the server console and login without any problems and ping out from the SCO server. The ping from the SCO server does not work. The SCO server at the console seems to be working OK, whats the problem ?
My solution has been to reboot the system when this problem occurs.
I have notice arp errors in the /var/adm/messages file. Could this be related ? Messages printed below:

WARNING: arp: info overwritten for 172.25.12.76 by 0:0:b5:1c:34:5b
Wed Feb 6 10:20:32 2002

Why is this server droping its TCP/IP stack ?

Please help. I have no other support for this SCO server.

 

[stanhubble]

Is 172.25.12.76 the ip of your sco server?
what version of sco?

when the tcp is "hung" what do the following commands respond with:
netstat -nr
netstat -ni
arp -na
ping -n SOME.IP.ONYOUR.NETWORK

this should give an indication of what is happening.
my suspicion is either the route table and/or something is grabbing your ip address.
or it could just be a flakey network card.

stan hubble

 

[pcunix]

arp info overwritten means that some other machine is using that ip address. If that's not the address of the server, that's relatively unimportant.

As Stan said, dropping the network is often the sign of a flakey card. I'll add that it is also the sign of a CHEAP card- why people put $12.00 nics in $3,000 servers is beyond my ken.

Finally, it may be that your routing table is being changed by rip packets. Kill the routed daemon and edit /etc/tcp to keep it from restarting if "netstat -rn" doesn't show what you expect.

Tony Lawrence
SCO Unix/Linux Resources

http://pcunix.com

tony@pcunix.com

 

[pcunix]

You know, when I said that "arp info overwritten" is relatively unimportant, I meant in the context of this problem.

It does mean that you have two or more machines trying to claim the same address, and that will cause problems for those machines..

Tony Lawrence
SCO Unix/Linux Resources

http://pcunix.com

tony@pcunix.com

 

[wpdavids]

Thanks every one I killed the routed process and noticed improvements. First off the default gateway is not disappearing from the SCO server's routing table. I am just waiting to see if it will stay connected to the network and if the network group notices less broadcasts from this SCO server.

This would solve alot of support problems.

Does someone know how to stop the routed process from starting on boot ? I know it starts in /etc/tcp, but what needs to be commented out ?

 

[pcunix]

The line that says

/etc/routed &
Tony Lawrence
SCO Unix/Linux Resources

http://pcunix.com

tony@pcunix.com