That is a technical term called flapping. We had some early model Cisco 3500 XL switches that experienced that problem with every gigabit NIC we tested. Cisco eventually admitted to the problem and replaced all of our 3500 XLswitches.
Here is the show version information from our failing switches:
32K bytes of flash-simulated non-volatile configuration memory.
Base ethernet MAC Address: 00:07:EB:95:40:C0
Motherboard assembly number: 73-3904-12
Power supply part number: 34-0851-02
Motherboard serial number: Removed
Power supply serial number: Removed
Model revision number: A0
Motherboard revision number: A0
Model number: WS-C3524-XL-EN
System serial number: FAB0545U3SB
Configuration register is 0xF
I dug a bit further and found a few references to the Cisco firmware revision about A0. My switch does have the A0 revision so I guess that is the problem. Is there a firmware upgrade or does the whole switch need to be replaced?
If your Cisco 3500's are A0/A0 like Droid said, it's the switch. We are fighting the same issue. I am trying to get cisco to replace all 90 of our switches, but they're still acting like they know nothing about it.
This seems to be an ongoing problem for many users of the Intel Pro/1000 MT NIC embedded on Dell's motherboards. After extensive testing and configuring of our clients switch (not a Cisco) and their NICs, the issue was isolated to the NIC. The chip that Dell uses from Intel is greatly modified from the original Intel chip for that NIC.
Our solution: We disabled the ethernet controller in the BIOS and replaced the NICS with (any brand)standalone PCI NICS.
The performance has been absolutely flawless. While being a major supporter of Dell equipment, the embedded Intel Pro/1000 MT created to many conflicts for us to recommend its usage.
Beta, What you experienced may be the case sometimes. In our situation, the Cisco 3500s in question will not connect to any gigabit card at anything but 10 meg. We have tried standard Intel Pro/1000 add-ons, and even tried to interconnect the 3500 to a Cisco 2950 switch. Both of those failed.
I agree the Cisco switch creates an issue when connecting at a gigabit (or attempting to). I also suggest that the Intel embedded NIC (Intel Pro1000/MT) creates its own issues when establishing connectivity on any level. We tested it at different speeds with different switches and found its performance to be the same, intermittent). I believe the combination of issues has people looking at one vendor versus another when both the Cisco 3500 as well as the embedded Intel Pro 1000/MT have performance faults all their own.
My recommendation would be if Cisco switches are involved: have them replaced by Cisco (as they recognized their problem with the STP used by the switch)
If just the Intel Pro 1000/MT on the motherboard is present and no Cisco switch: my original recommendation is working seamlessly across the board for our clients.
I was having the same problem and talked to Dell no solution, talked to Intel no solution. Had our Cisco person check packets still did not solve the problem. Then I went in to Network Connections --> Intel --> Configure --> Advanced--> and changed flow control to Off. Have not had any problems since.
Let me know if this works for you guys because we are testing this in a large environment.
As I posted previously, you may want to purchase a small pack of basic 10/100 nics (any brand), unless you have a larger network. Inwhich case our solution may be a little cumbersome.
Disabling the onboard NIC in the BIOS and installing new standalone cards was the only solid solution that worked for us. We did extensive testing to troubleshoot and isolate the problem. This solution may also work for you.
I been reading this thread due to similar problems we encountered with the GX2600 in our firast pilot post in Bruxelles, as senior networking specialist I was assigned to sort out the case.
The situation:
We bought a first batch of 300 GX260's, all were washed with a NT4.0 WS SP6a, about 30 of the machines went to Bruxelles, the were hooked up to a Cisco Catalyst WS2950G-48-EI switch, which should provide a more current interface port then the mentioned 3500XL series (more about that later).. Some of the GX260's connected fine, some only after e repeated boot, some after a little delay and some, specially one not at all.. no flapping here, no total refusal to establish a link at all.
The path sofar:
I acquired a few GX260's from our stock-room and hooked them up to a similar 2950, a (yep 3500 wit A0 rev mainboard) 3548G a 4006 and on 3 none Cisco platforms, on ALL save a very cheap (<30US home switch (Edimax)) the GX260's had problems establiching a link flawlessly, reliable or at all.. so blaming the 3500 series catalyst is IMHO rather shortsighted if not erroneous in finding the real source of the problem... which I hope can be determined after I get the one from Bruxelles that refused all the time. The problem is at least a bit more widespread then 'just the ole 3500XL' and definately more complex.
I tried all combinations of setting that could remotely have to do with the NIC and switch combinations.. different ports, duplex, speed, flowcontrol, burst, cable etc etc modes on all the formentioned switch platforms, I flashed the BIOS to A06, changed IRQ settings etc in the BIOS (talk about that people at Dell, is it really smart to combine the IRQ for USB and the Networkadapter if I change one the other keeps pace and the other way around, would be ever so nice to put the NIC on a possible IRQ ALONE?) and even put XP pro on the machines stead of the NT image, to no avail, still the results are random and performance is not up to current technology standard,
Just on the very dumb and cheap little Edimax all the GX260's pinged eachother just fine contineously and a 270MB FTp session went over in reasonable performance.
What's next:
I will wait for the machine from the embassy in Bruxelles to arrive and take that as the die hard case to get ready and rumble, plans are to try any setting combination (made a matrix for that) with severla OS and driver combinations, on all the above mentioned Cisco switches) My first (chaep shot) will be to install RH Linux on the 'black box' from Bruxelles, I got a hope that may just prove it may not be hardware dependent at all...
Will keep ya informed, sorry for the typo's do this forum as a quicky in between..
DELL-Cody
2.2K Posts
0
July 29th, 2003 13:00
droid256
73 Posts
0
July 29th, 2003 14:00
That is a technical term called flapping. We had some early model Cisco 3500 XL switches that experienced that problem with every gigabit NIC we tested. Cisco eventually admitted to the problem and replaced all of our 3500 XLswitches.
Here is the show version information from our failing switches:
32K bytes of flash-simulated non-volatile configuration memory.
Base ethernet MAC Address: 00:07:EB:95:40:C0
Motherboard assembly number: 73-3904-12
Power supply part number: 34-0851-02
Motherboard serial number: Removed
Power supply serial number: Removed
Model revision number: A0
Motherboard revision number: A0
Model number: WS-C3524-XL-EN
System serial number: FAB0545U3SB
Configuration register is 0xF
Our new working switches are Model Revision # S0.
Droid
123harry
3 Posts
0
July 29th, 2003 21:00
Thanks droid256,
I dug a bit further and found a few references to the Cisco firmware revision about A0. My switch does have the A0 revision so I guess that is the problem. Is there a firmware upgrade or does the whole switch need to be replaced?
droid256
73 Posts
0
July 30th, 2003 13:00
Cisco replace our switches.
Droid
ccclub
1 Message
0
July 31st, 2003 02:00
droid256
73 Posts
0
July 31st, 2003 13:00
Griff76
13 Posts
0
August 4th, 2003 15:00
MBarry
1 Message
0
August 8th, 2003 13:00
MB
beta2
6 Posts
0
August 12th, 2003 12:00
This seems to be an ongoing problem for many users of the Intel Pro/1000 MT NIC embedded on Dell's motherboards. After extensive testing and configuring of our clients switch (not a Cisco) and their NICs, the issue was isolated to the NIC. The chip that Dell uses from Intel is greatly modified from the original Intel chip for that NIC.
Our solution: We disabled the ethernet controller in the BIOS and replaced the NICS with (any brand)standalone PCI NICS.
The performance has been absolutely flawless. While being a major supporter of Dell equipment, the embedded Intel Pro/1000 MT created to many conflicts for us to recommend its usage.
Griff76
13 Posts
0
August 12th, 2003 17:00
Beta, What you experienced may be the case sometimes. In our situation, the Cisco 3500s in question will not connect to any gigabit card at anything but 10 meg. We have tried standard Intel Pro/1000 add-ons, and even tried to interconnect the 3500 to a Cisco 2950 switch. Both of those failed.
beta2
6 Posts
0
August 12th, 2003 21:00
Griff,
I agree the Cisco switch creates an issue when connecting at a gigabit (or attempting to). I also suggest that the Intel embedded NIC (Intel Pro1000/MT) creates its own issues when establishing connectivity on any level. We tested it at different speeds with different switches and found its performance to be the same, intermittent). I believe the combination of issues has people looking at one vendor versus another when both the Cisco 3500 as well as the embedded Intel Pro 1000/MT have performance faults all their own.
My recommendation would be if Cisco switches are involved: have them replaced by Cisco (as they recognized their problem with the STP used by the switch)
If just the Intel Pro 1000/MT on the motherboard is present and no Cisco switch: my original recommendation is working seamlessly across the board for our clients.
Thanks for the reply!
Beta
ROB5
1 Message
0
August 13th, 2003 18:00
I was having the same problem and talked to Dell no solution, talked to Intel no solution. Had our Cisco person check packets still did not solve the problem. Then I went in to Network Connections --> Intel --> Configure --> Advanced--> and changed flow control to Off. Have not had any problems since.
Let me know if this works for you guys because we are testing this in a large environment.
JustinBaiocchi
5 Posts
0
August 13th, 2003 21:00
beta2
6 Posts
0
August 18th, 2003 12:00
Justin,
As I posted previously, you may want to purchase a small pack of basic 10/100 nics (any brand), unless you have a larger network. Inwhich case our solution may be a little cumbersome.
Disabling the onboard NIC in the BIOS and installing new standalone cards was the only solid solution that worked for us. We did extensive testing to troubleshoot and isolate the problem. This solution may also work for you.
Beta2
BLacq
4 Posts
0
September 4th, 2003 10:00
I been reading this thread due to similar problems we encountered with the GX2600 in our firast pilot post in Bruxelles, as senior networking specialist I was assigned to sort out the case.
The situation:
We bought a first batch of 300 GX260's, all were washed with a NT4.0 WS SP6a, about 30 of the machines went to Bruxelles, the were hooked up to a Cisco Catalyst WS2950G-48-EI switch, which should provide a more current interface port then the mentioned 3500XL series (more about that later).. Some of the GX260's connected fine, some only after e repeated boot, some after a little delay and some, specially one not at all.. no flapping here, no total refusal to establish a link at all.
The path sofar:
I acquired a few GX260's from our stock-room and hooked them up to a similar 2950, a (yep 3500 wit A0 rev mainboard) 3548G a 4006 and on 3 none Cisco platforms, on ALL save a very cheap (<30US home switch (Edimax)) the GX260's had problems establiching a link flawlessly, reliable or at all.. so blaming the 3500 series catalyst is IMHO rather shortsighted if not erroneous in finding the real source of the problem... which I hope can be determined after I get the one from Bruxelles that refused all the time. The problem is at least a bit more widespread then 'just the ole 3500XL' and definately more complex.
I tried all combinations of setting that could remotely have to do with the NIC and switch combinations.. different ports, duplex, speed, flowcontrol, burst, cable etc etc modes on all the formentioned switch platforms, I flashed the BIOS to A06, changed IRQ settings etc in the BIOS (talk about that people at Dell, is it really smart to combine the IRQ for USB and the Networkadapter if I change one the other keeps pace and the other way around, would be ever so nice to put the NIC on a possible IRQ ALONE?) and even put XP pro on the machines stead of the NT image, to no avail, still the results are random and performance is not up to current technology standard,
Just on the very dumb and cheap little Edimax all the GX260's pinged eachother just fine contineously and a 270MB FTp session went over in reasonable performance.
What's next:
I will wait for the machine from the embassy in Bruxelles to arrive and take that as the die hard case to get ready and rumble, plans are to try any setting combination (made a matrix for that) with severla OS and driver combinations, on all the above mentioned Cisco switches) My first (chaep shot) will be to install RH Linux on the 'black box' from Bruxelles, I got a hope that may just prove it may not be hardware dependent at all...
Will keep ya informed, sorry for the typo's do this forum as a quicky in between..
BLacq