One branch hangs

[MiniYC]

We have 3 different branches connected by T1 to the main brnch. One of the branch would loose its connection to the SCO Unix server but still has the network connection to the other servers to the main branch. A computer from that branch cannot ping the Unix box, but the Server (w2k) in that branch can. T1 line been tested nothing wrong.

We end up telling everyone, including the other branch to log out and we reboot the Unix server. That seems to fix the problem. We can afford to always reboot the server.

Anyone have any idea what could be wrong????

 

[stanhubble]

We probably need more information (version of sco, is there any subnetting going on, etc) but this sounds like a routing problem. routed can learn a route that can confuse everything else. unless you know for a fact that you need dynamic routing make sure routed does not run when the tcp stack is started.

check the output of netstat -nr before and after the condition occurs.

 

[MiniYC]

Im new with Unix.....

We do have subnetting.
How do I find out the version of SCO?

 

[stanhubble]

for version of sco(presuming openserver):

uname -X

for routing information:

netstat -nr

 

[MiniYC]

Version 5.0.5

 

[stanhubble]

In openserver 5.0.5 routed starts by default.
assuming your routes are going to be static....
(this needs to be done before you experience the problems)

see if routed is running. ( ps -ef | grep routed )
if it is , then get the PID from the previous command and "kill" it.

netstat -nr should show you the routes to your branch sites if they are on different subnets.

ICMP can still add routes to the routing table but the usual culprit is routed.

if this "cures" the problem, then you will need to either edit the /etc/tcp file to prevent routed from starting when you reboot and/or remove the routed program from the system (i tend just to rename it).

hth
stan

just to be clear:
the sco server is at the main site.
w2k server(s) are at branch sites.
clients at a branch site loose contact with sco server.

 

[MiniYC]

Yes ...
SCO at the main site and w2k servers at the branches. Clients at a branch loose their connection to the SCO but can network connections to the other servers to the main branch are okay.

Then w2k server at a branch can ping the SCO server but the clients cannot.

 

[MiniYC]

I kill routed, still had one branch go down, actually twice yesterday.

Any other ideas what could be wrong...

 

[stanhubble]

Is it the same branch that keeps dropping?

what did the route table look like before the problem
and after the problem? (netstat -nr)

Is there a difference in the addressing schemes between the w2k server and the clients?
Is there a difference in address resolution between the w2k server and the clients?

 

[MiniYC]

Yes it's the same branch. I didnt get a chance to check the route table.

addressing schemes and resolution are the same.

 

[stanhubble]

hmmm....
the thing that bothers me is that the w2k server still pings ok, which says the sco system still knows how to get back to it.(static route set for that host rather than a subnet??)

I still think you will find some clues in the routing table.

I won't be around for a week or more, so maybe someone else has some insight.

good luck
stan