Mystifying switch problem

[milleron]

I use a couple of gigabit switches to connect my desktops and a gigabit
Linkstation NAS in different parts of the house.

A week or two ago, I started losing connection to the Internet, the
router, and the NAS from the desktop computers. The computers' Tray
icon still showed a "1.0 Gbs" connection, but I could not communicate.
It turned out that all that was necessary to re-establish the connection
was to power-cycle the switch in the room with the two desktops.
Unfortunately, the connection would be lost within two minutes requiring
the process to be completed ad infinitum. Bypassing the switch and
connecting the desktops directly to the other switch worked fine.
CONCLUSION: the upstairs switch was bad.

I asked for advice on this forum about what to replace it with, and
followed that advice. Today the new Netgear Prosafe GS108 gigabit
switch came, and it installed nicely. It worked for a few minutes, and
then exactly the same problem recurred. Power-cycling the new switch
restores the connection exactly like the former switch, but this
power-cycling has to be repeated every few minutes.

CONCLUSION 2: The switches are probably not bad because the chances
that the old switch and the brand new one would fail in exactly the same
way are vanishingly small.

CONCLUSION 3: The other hardware devices in the network (other gigabit
switch, router, and NAS) are OK because they all work fine with the
upstairs switch removed from the network.

I changed the switch's power supply connection from an outlet on the UPS
that's not backed up to one that is backed up by the UPS battery --> no
change. Of course, the power transformer, itself, was swapped along
with the new switch. The UPS is new, by the way -- installed about
three weeks before the problems started.

I tried power-cycling the other switch (to which the failing switch is
connected), but that won't re-establish the connection.

I swapped out the Ethernet cable that goes from the failing switch to
the wall jack: no joy.

QUESTION: What could be causing the switch to which the desktops are
connected to lose its connection to the rest of the network repeatedly
and continuously?????

 

[Jack \(MVP-Networking\).]

Hi
I hope that you understand that not all matters that involve Networking (as
well as other topics) can be solved via a communication form like a
newsgroup, or over the Internet in general. Some times there is No other
choice but to get a pro to come and actually evaluate and fix the problems
in person.
That said, could be that some thing is wrong with the topology, and or, the
cables arrangement that lead to the switch in question.
At this point to I doubt that it can be resolved without precise map of all
the connection and the general IP scheme of the Network.
May be that you can install Qcheck and run it between the concerned computer
within the few minutes that the Network works.
When the connection is lost do you lose the IPs, can you ping, what are the
more precise Network variables that stop function, on all computers.
The term lose the connection is too broad and does not really provide any
focused information.
http://www.ezlan.net/faq.html#qcheck
Jack (MVP-Networking).

 

[milleron]

Losing the connection in the context of my post meant that the
computer ran fine, but couldn't see the LAN or the internet any
longer. It did not lose it's router-assigned IP, but, of course, it
was impossible to ping or run tracert, and it was impossible to
release and then reobtain another IP. The switch was "locked" with
flashing LEDs.

I bought a THIRD switch, this one a SOHO variety by D-Link, and
exactly the same problem recurred. This I considered proof positive
that the switches were locking up because of a device connected to
them and not because they were faulty. I knew that both computers
connected to the switch worked fine when connected directly to the
next switch down the line, so that left only the Netgear USB print
server. In all my troubleshooting, I'd never thought to disconnect
that little print server. I unplugged it, and the D-Link switch has
worked fine for over three hours now. It remains to be seen if
power-cycling the print server allows it to be connected without again
locking up the switch.

The topology, by the way, was as follows (if the ASCII characters
don't mess up in transmission). Everything beyond the router is
gigabit capable EXCEPT the little print server. The switch depicted
in all-caps was the one that continuously and repeatedly locked up and
required power-cycling.

Cable modem
|
Vonage adapter
|
Netgear wireless router (10/100)
|
Netgear gigabit switch -- gigabit NAS
| |
5e |__cat 5E__living-room and kitchen jacks
|
NETGEAR GIGABIT SWITCH in study -- print server
| |
5e computer 2
|
computer 1

The vonage adapter assigns an IP to the router which, in turn, assigns
IPs to all the devices on the network, EXCEPT the little print server
which has a STATIC IP. All the devices are 192.168.1.1 - 10

 

[Jack \(MVP-Networking\).]

Hi
Disconnect the one computer from the Network and plug the Printer server
directly to the computer's NIC.
If it can converse with the computer give it a static IP that is of the same
subnet as your Network but out of the DHCP range.
If it works with this configuration with the computer try it to the Network
again.
In general I do not use 3rd party printer servers, they do tend to be
"quirky". Given that Network Ready Printers are not that expensive any more,
any one of them would be better then 3rd party "gizmo".
Jack (MVP-Networking).

 

[milleron]

Thanks VERY much for your replies. I've been into computer hardware
for almost twenty years, and I've been building my own computers for 8
years, but this is the most difficult troubleshooting riddle I've ever
faced.

Turns out it wasn't the print server. I unplugged that thing and it
worked for several hours, but it finally lost the connection again.
Since the UPS was a relatively new addition, I took that out and
replaced it with the old one. Again, we went for about 5 or 6 hours,
but the connection was then lost again.

This network has been working flawlessly for three years, albeit with
conventional 10/100 switches until about a year ago when I swapped
them out for the gigabit switches. Then the gigabit switches worked
flawlessly for a year.

My next troubleshooting step is to dismantle the part of the LAN in my
study and just run the main computer connected directly (via a jack
and cable to the room with the modem and NAS).

Again, thanks for helping me think this through.

Ron