Catalyst 2900 trunk problem...

[Connie]

I have a few Catalyst 2900 switches trunking to a Catalyst 5500 that has this problem: On the 2900 trunk ports, the LEDs are flashing amber and green all the time. Switch performance is affected, but I don't know what the problem can be, as I have a few exact same 2900's connected to the same blade on the 5500 without this problem!

We do not have spanning tree enabled on our LAN, and all trunk ports are setup for ISL, and nailed to 100Mbps full duplex.

I am also picking up errors on the 2900 trunk ports in the form of giants, but there are no giants on the 5500 trunk ports...

 

[wybnormal]

Well.. first I question not having spanning tree enabled. That my friend is playing with fire. Has anyone certified the cabling used for the trunks to be CAT5? and I mean using something like a Pentascanner 350 to actually work the cable.. not just a wiring checker. Bad cabling has bitten more people then almost anything else.

Are all the switches on the same level of code? Which level of code is the Cat5K on and which blade are you using? some of the older blades have problems with trunking per the various readme files. The older 12 port 13A is one that can have serious problems depending on the 5K code level.

MikeS Find me at

http://www.packetattack.com

"The trouble with giving up civil rights is that you never get them back"

 

[Connie]

Thanks for the reply, Mike!

To answer a few points you raised: Should you enable spanning tree only if you have more than one link to a specific device? We are running a pure star LAN, with point-to-point trunks to all switches.

The 5505 and the 2900 are in the same wiring closet, with a 0.5m cross-over flylead connecting them. This flylead has been certified by our local cabling contractor, and by myself (with mentioned Petascanner +).

Model and version of software/hardware in the 5505:
WS-X5225R Hw : 3.3, Fw : 4.3(1), Sw : 5.5(9).

Version of software on the 2900: 11.2(8.1)SA6.

I have just noticed today that throughput on that switch is poor. Takes about 100 minutes to copy a 400MB file to another PC on the same switch - both nailed @ 100MB Full duplex.

 

[wybnormal]

Spanning tree is a safty net in case a rogue switch is installed or a cable.. some other protocols use it like Etherchannel to help manage things.

Good news on the cable.. Pentascanner+ is fine tester.

The code on the 2900 is very old.. and is known to have some issues trunking. You should really consider upgrading it to a newer code. I have 12 on my mine and it's very stable.. so far ;-) The code on the Cat is good at 5.5.. thats pretty recent and I have had good luck with it on several Cats.

An interesting trick to try.. nail the 100 to 100 HALF and see what happens. If it improves then look REAL hard at a cabling issue.. is the cabling by anything electric? power main? I got tagged once due to a cell-site across the parking lot was nailing a cable run with excessive power.

The best thing is if per chance you have access to a sniffer.. get a trace on the link and see just what is happening. retransmissions? delays at the switch? fragmented packets?

MikeS Find me at

http://www.packetattack.com

"The trouble with giving up civil rights is that you never get them back"

 

[Connie]

Mike, I think I have solved the problem! I enabled STP on the 2900, and things seem to have improved quite drastically. The trunk port's status LED stopped alternating amber and green (no more spanning-tree issues?), but the real proof is in the pudding:

It now takes 6 minutes to transfer a 762MB file between the same two PCs that managed 100 minutes for a 400MB file! The same 400MB file took less than 5 minutes from one PC to the other.

Would the solution really be as simple as enabling STP on the 2900 - it just sounds too good to be true?

Connie

 

[wybnormal]

Without being at the console, I'm fishing. I will offer this advice. Make the Cat5000 the root bridge. You can hard code this from 32768 to 8192( I think these are the numbers). This is to prevent one of the 2900's to think *its* the root bridge.

If I had to guess, I would say it's an interaction between VLAN1(managementVLAN) and Spanning Tree. I still say you should upgrade the code

MikeS


Here are few articles.

Troubelshooting STP

http://www.cisco.com/warp/public/473/16.html#4d

Configuring SPanning Tree

http://www.cisco.com/univercd/cc/td/doc/product/lan/cat6000/sft_6_1/configgd/spantree.htm

ISL Trunking on the 5000

http://www.cisco.com/warp/public/793/lan_switching/2.html

Find me at

http://www.packetattack.com

"The trouble with giving up civil rights is that you never get them back"