Boggling network problems - anyone a master network troubleshooter?
 

I have a very strange problem occuring on our company's internal network. If anyone can give me ideas as to why this is happening, I would officially proclaim you a demi-god.

What I want:

to have regular access from my Redhat Linux 9 machine @ 89.80.89.28 to another Redhat Linux 9 machine @ 10.5.220.4

This isn't happening. Instead, the connection is very intermittent. It will be fine for a while, but after about 2 or 3 minutes of inactivity, it dies. At one moment, I can ping it and it responds in < 2 ms. However, after a while, I will ping it, and it won't respond for 3-10 minutes! It will then begin responding normally (< 2 ms) again.

What could be going on? Here is the network topology-

89.80.89.28 ---> eth switch ---> 89.80.89.100[cisco router]10.5.1.40 ---> switch/hub --->10.5.220.40

Unfortunately, all the equipment after the router is managed by our MIS department, and I am not sure how it is all configured.

Other notes:

89.80.89.19 (HPUX 11.11) can ping 10.5.220.40 anytime without trouble, and it doesn't even have an active connection (I have an NFS connection between the two Linux machines)

89.80.89.26 (HPUX) can ping 10.5.220.40 anytime.
89.80.89.55 (WinNT) can ping 10.5.220.40 anytime.

They can ping back. 10.5.220.40 can ping them, but not the linux machine (89.28)

Other machines on the 10.5.x.x network that the Linux machine can't ping, but others can:
10.5.1.5 (our router to the outside world)
10.5.1.8 (our DNS server)
several others

More strange stuff:

The Linux machine can ping 10.5.1.249 anytime and get a reply, while at the same time not getting a reply from the others mentioned above. As far as I know, they are on the same network (all the machines there have subnet mask 255.255.0.0)

So, the final question:

How come I can see some machines and not others right next to it?
Our router doesn't seem to be a problem, as other machines on our side can see anyone on the other side.

Ideas? Suggestions? Do you want more info?

Thanks

I think there is something missing. I am wondering what the distances are in your network topology. Sometimes if a computer has a very long Ethernet link to the next powered device a connection can seem very intermittent.

I am also wondering what your network capacity is for your router and switches. And to be sure, find out if the networking device by 10.5.220.40 is a switch or a hub, and if it is a hub - find out if it is an active or a passive hub.

Passive hubs merely extend a direct wired connection between two powered devices.

you wouldint happen to have anything monitoring the bandwith useage on your hub/routers

Thanks for your ideas. I don't think we have anything monitoring the bandwidth on our router. It is not managed. It is an old Cisco 1520 series router running at 10 Mbit/s There isn't much traffic over it. The two networks generally do their own thing. We use MIS for email and internet access, as well as access to the VAX system.

I'll check to see if bandwidth is being limited.

As for distances in the network - one connection, by the 100 BaseT standard, can be 100Meters, right? The diameter of the whole area in which we are working is about 75 meters. But I can see how it might be a distance problem.

I suspect it isn't for these reasons:

-The machine that doesn't work is closer to our router than most of those that do work. Our entire network goes through the router. After the router, the distance is the same for everyone. Yet the Linux machine is the only one having trouble reaching the other side.

I am wondering if something could be filtering packets on the application level, or by IP address. Is it possible that routing tables are messed up somewhere? Is it possible that the Linux machine doesn't like the IP address of the other Linux machine? I suppose I might have to look into that.

Any other ideas? Thanks!

How is the connection between the HPUX, WinNT and the Redhat at (89.28) machines on the local subnet. What I'm really asking is if the (89.29) connection to the switch is causing problems. Some switches are configurable, 10 vs 100 mb, full vs half duplex. There could be a problem auto-configuring the connection. Or the switch input is configured and not the right way. Also, there could be problems with the connector at either end or the cable between the switch and the Red Hat machine. If the connections on the local subnet are OK, then you can eliminate these possibilities. Then, you can concentrate beyond the switch.