[AySz88]

I've been going crazy trying to diagnose what's going on with a wired network. There's four routers (three being used as switches, two physically inaccessible in a wall) and about a dozen computers and devices. The setup is currently like this:

Main ISP modem/router: 192.168.15.1, hosts the DHCP server for the whole network
Wireless router being used as an access point: 192.168.15.4, connected to the main (...1) router via its LAN port, with DHCP disabled
Computer A connected via LAN to the wireless router
Computer B connected via Wi-Fi to the wireless router
Wired router being used as a switch: 192.168.15.2, connected to the wireless (...4) AP, again all through LAN ports with DHCP disabled
Wired router also being used as a switch: 192.168.15.3, connected to the other switch (...2), ditto
Computers X connected to either the .3 or the .4 switch

(So the routers are daisy-chained 1..4..2..3 - sorry about them being out of order.)

All the computers are Windows XP or Windows 7 and almost surely don't have any weird configuration changes compared to when they were new and out-of-the-box.

The affected computers have websites that get "stuck" loading - the symptoms are as if the computers intermittently disconnect. To diagnose things, I have been running pings with lots of large packets, like "ping 192.168.15.1 -t -l 256 -w 100". The problem shows up only very intermittently if I ping using the default 32-byte packets.

Computer A is able to ping all the routers and the Internet without significant packet loss. Ditto for computer B.
Computer X suffers very high packet loss (about 35% dropped) to all the routers (including the switch it is directly connected to) and to servers on the Internet.

I can't figure out why A is able to ping fine to both switches, but computers connected to the switches cannot ping back out reliably. The computer that can't ping the switch it's connected to is really bizarre. I also am puzzled about what would make small packets okay but large packets not - line noise? Miswires? Collisions?

[edit] It seems that not all the computers on the switches have this problem...still trying to figure out what the pattern is.
[edit2] And now one computer connected via Wi-Fi is having the same problems, but not the others that are via Wi-Fi; this is too weird. :/
[edit3] I think the Wi-Fi thing may be a different problem...disregard for now, I suppose.

Any suggestions for directions or additional diagnostics at this point?

Edited by AySz88, 12 March 2011 - 05:18 PM.

[Advertisements]

Sounds like a DUPLEX mismatch. Probably X is set with Full Duplex and the switch is set with AutoNegotiate. (This results in the switch setting itself to Half-Duplex.) If you do a speed check
http://www.speedtest.net/
you will probably see that uploads from X are pretty normal but that downloads are much slower than expected.

Ron

[RKinner]

Sounds like a DUPLEX mismatch. Probably X is set with Full Duplex and the switch is set with AutoNegotiate. (This results in the switch setting itself to Half-Duplex.) If you do a speed check
http://www.speedtest.net/
you will probably see that uploads from X are pretty normal but that downloads are much slower than expected.

Ron

[dsenette]

any chance of undaisy chaining all of these things? can you just plug everything into the main ISP router?

i concur with the possibility of a duplex missmatch.

also, depending on make/model of the equipment, there could be some kind of odd custom config issue.

[dsenette]

any chance of undaisy chaining all of these things? can you just plug everything into the main ISP router?

i concur with the possibility of a duplex missmatch.

also, depending on make/model of the equipment, there could be some kind of odd custom config issue.

[AySz88]

An update... Replaced the dying wi-fi router with a new one, but the other problems are still there. The computers are all set for auto-negotiate, but the switches don't have any place to see or set their duplex settings. I'm going to replace those too and see if the problem disappears....

[RKinner]

I was thinking the Full Duplex problem would be on the PC X not the switches. Do you know how to check that?

Ron

[AySz88]

With the new switches, most of the computers now either don't have the problem, or got rid of the problem after a reconnect or reboot. There's one computer that is so bad that it refuses to recognize a connection long enough to grab an IP address. (If I give it a static IP, it just exhibits the original pinging problems.) Oddly enough, if I put together a "network" just for it (i.e. with the stuff I replaced), it seems to work fine. I'm suspicious about that computer - it's been exposed to all sorts of kitchen nasties - so I'm going to replace the Ethernet adapter on that one computer to try to make sure I'm not confounding another different problem there.

I ran a speed test on another computer while it was having problems (not the really-misbehaving one) and it was uploads that were slow, not downloads.

I was thinking the Full Duplex problem would be on the PC X not the switches. Do you know how to check that?

Ron

Yeah; the computers connected to the switches is auto-negotiate. I should be clear that there's multiple things in the place of "computer X", and each of them were intermittently having the problem.

[RKinner]

Might be a bad NIC.

We are also seeing a fake NIC driver being installed by malware. Once the system is clean we have to go into Device Manager then find the Network Adapters and uninstall each item under Network Adapter. Then Reboot. Windows will find the good one and reinstall it.

Ron

[AySz88]

Just updating this for anyone that happens to run into the same problem again - I bought a Fluke MicroScanner to check the wiring in the walls (thankfully, it can be a business expense). The problem was that the contractor who put in the ethernet cables in the walls of our apartment managed to wire almost *every* endpoint incorrectly, but in a consistent way. (They swapped the brown-white wire with the orange-white wire. So instead of 1-2 and 7-8 being twisted pairs, they made 1-8 and 2-7 twisted, allowing interference to occur between them.) Going to have to get them to redo every endpoint. :/

Edited by AySz88, 21 May 2011 - 01:32 PM.