UNIX for Advanced & Expert Users

While installing a firewall, I was pinging the interface from SCO 5.0.6 Openserver box, while no response, I hit "DEL" to cancel, but no cancel. Then all of a sudden I get BOO-KOO traffic lights on HUB and Switch.... Then a kernel trap error. System froze... Proceeded with a cold boot. Now I have no trace of the kernel trap, no visible trace of any problems other than with any network services, such as telnet, ftp, rlogin, etc, etc.... anything over the NIC I get up to 2 minute delays on login.
Even running netstat takes like 5 minutes.

I've checked every place I can think of. An identical system with different address has no problems what so ever.

Sometimes in the AM it delays for a while then it get back to normal about mid-morning. Then after no changes, the next morning everything is fine, then it goes back on delay in the afternoon. Only other thing I can offer is that in syslog, I see a few rlogind's posting unable to find user's info in utmp.

This system is very basic setup. Using NFS for one system, getty for serial lines (most use), then rlogins for laptop users, and FTP to do files over 1.3 MB.

Being as SCO is being the IX BIOTCH these days and requiring $599 per incident I was wondering if you's guys can bail me out here?

Thanks,
Brandon.

I have replaced the NIC and still no help.

I come in and rlogin is fine but FTP is the issue.
inetd.conf is it's default config right now. FTPaccess is blank other than root.

ANYONE?!?! Any ideas where to dig? I'm looking at streams and signal send/recv info now.

Kernel parameters show that my NSTREAM (stream head structures) is up somewhere around 13000 when it says normal is about 16.

Is this a BINGO flag?

A comment like "FTP is the issue." is only a tiny step above "Sometimes is wrong with my box." I hate to tell you, but not all FTP issues have the same solution.

Very few people here use SCO. Combine this with your sparse error reports and we have nothing to go on.

In general, I debug network problems from the network. When you got those high traffic lights, you needed to use a sniffer to see what the traffic is. Something like "We are getting a flood of UDP datagrams for DNS from ip address a.b.c.d to ip address z.y.x.v." gets you much closer to a solution than "BOO-KOO traffic lights on HUB and Switch".

Same deal with ftp. Watch the conversation on a sniffer. See where the protocol breaks down. (Assuming it does break down. It's not even clear if you want ftp to succeed or be blocked.)

Well, it's hard to give all the information on a post.

Thanks for the reply, by the way.

As the systems sits right now, this is a small private network of less than 50 XP pc's with 2 unix boxes. No DNS is used. Everything is by IP address.

With normal traffic flow now and everything appearing to be back to normal, my issue is that all network related services are skitish. The only ones we use are rlogin, ftp and NFS. 90% of those 50 ish PC's are using serial tty connections to connect to the unix boxes. So this leaves me with 3 ideas:

1. Some inet related file is corrupt,
2. Some services has been disabled,
3. Or inet services have lost priority to something low like mail.

Can you point me some dirs/files to check these ideas, mostly with #1 or #3.....???

Thanks

You might want to check any differences in setup of the two UNIX systems.

Take a look at Routing - especially since you noted "An identical system with different address has no problems what so ever."

To help us assist you, you might want to let us know the IP of the two UNIX boxes, the default route on each, if there is a router between them...but first check out the link. It might help you realize your problem.

"my issue is that all network related services are skitish."

*sigh* No one can help you with a problem report like that.