I have 5 boxes, all with internal 10.0.0.0 addresses. The set up:

Box 1 is Win2K Server, IP 10.0.0.1
Box 2 is Win98, IP 10.0.0.2
Box 3 is Red Hat 9, IP 10.0.0.10

Net mask is 255.0.0.0 for everyone.

Boxes 2 and 3 are on the same hub (downstairs). Box 1 and the unnamed boxes are on the same hub, with a downlink to the other hub where Boxes 2 and 3 are. Everyone was living happy and content and all could connect as needed.

Then I went on vacation. I needed to borrow some Cat-5 for on-the-road connectivity, so I took the chunk from the hub downstairs to Box 3 (Linux) (this is a red herring!). Then I shut Boxes 2 and 3 down, which does not happen often.

So I come back, boot up Box 3, and no connection from Box 1 to Box 3. Oops, forgot the cable. Patch it in, and still no connection from Box 1 (using PuTTY). Figure Box 3 didn't boot right, and reboot it. Still no joy. Hook up a monitor, and start pinging all over the place. I can ping Box 2 (on the same hub), but not Box 1 (on the upstairs hub). I jump on Box 2, and I can ping both Box 3 (Linux, on the same hub) and Box 1 (Win2K, on the hub upstairs), so it can't be a hub problem.

In short, Box 2 (on the same hub as Linux) can ping and get pinged by everyone. Box 1 can ping everyone except Box 3. Box 3 can ping no one except Box 2, on its same hub, and can get pinged by no one except Box 2.

To further add to the mess, even though Box 2 can ping Box 3, it cannot open an http session with it, even though httpd is running on Box 3.

It seems like the routing table on Box 3 is hosed, or something?

TIA for any help!

I doubt if it is a routing problem as you then shouldn't be able to access box2. You can check with:
route -n

I'm more inclined to guess it is a firewall problem:
iptables -nL

Are you sure that your webserver is listening?
netstat -nlp

Thanks for your reply!

True about the routing, since all the boxes are on the same subnet if I can get to Box 2 then Box 1 shouldn't be a problem.

At any rate, route -n shows that routes to 10.0.0.0 use gateway 0.0.0.0 on interface eth0 (which is the only NIC in the box, and I'm not using IP aliasing).

iptables -nL shows three headings, Chain INPUT, chain FORWARD, chain OUTPUT, all with policies of accept. Under each of these there are what seem to be table headings (target, protocol, etc.) but no entries in the table. I've never configured iptables, so I guess it would make sense that there are no entries.

I believe the web server is listening, in httpd.conf I have configured Listen 80, and the netstat shows that the box is listening on 0.0.0.0:80 for TCP traffic.

Here is something interesting: when I entered wget http://127.0.0.1/aoe.php (which is a page in the directory root), I get this:

Connecting to 10.0.0.1:80...

Failed to connect (or some such), no route to 10.0.0.1.

Why would connecting to the loopback address get redirected to 10.0.0.1 when this machine has IP 10.0.0.10? And why would it say "no route to 10.0.0.1" when it clearly can find a route to 10.0.0.2? Weirdness...

And even more--why all these troubles when nothing more has been done other than to shut the box down for a few days? The last time it was shut down, nothing weird happened.

TIA for any help.

The only reason I can see it trying access 10.0.0.1 is if the loopbac interface isn't up - check with:
ifconfig eth0

Just to confirm - does your routing tbale look like this:

It almost looks like that, except that the mask for 10.0.0.0 is 255.0.0.0 (which is the net mask for all the boxes), and the metric for 0.0.0.0 is 0--which is not right, since it is clearly one hop to 10.0.0.1. It almost looks like the box thinks its IP is 10.0.0.1 instead of 10.0.0.10, although ifconfig confirms that eth0 has 10.0.0.10.

Any ideas?

Hi David,

Problem solved. I was installing XP on a box downstairs and borrowed the cable from Linux to get a temporary Internet connection. Guess what--it wouldn't connect! I couldn't ping the upstairs computers, but I could ping the other box on the same hub. So I switched the cable to a different port, and now I can ping everywhere!

When I came back from vacation and hooked the Linux box back up, the port I plugged into must have been different than the one I had been using as well as being flakey.

Anyway, thanks for all your help!

Well spotted - those problems are the difficult ones to find.

Glad to see you got it working.