Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations Mike Lewis on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

MLT Configuration on 3510

Status
Not open for further replies.

dhopp97

IS-IT--Management
Nov 3, 2005
3
US
I have 2 3510-24T switches that I am trying to configure MLT between to get an aggregate bandwidth of 2Gb/sec.

The MLT seems to work for failover (if I unplug one of the cables I lose no packets) but watching the traffic on the ports I only ever use one of the ports. It never seems to load balance over to the other port.

I have port 1 and 2 in an MLT Group. There are 3 vlans configured (1, 3, 5). I can access any of the vlans from either switch, but no matter which one I use, traffic is only sent accross one of the links.

Looking at the spanning tree configuration both ports are in "forwarding" state.

Any ideas?

Thanks,

--Dennis
 
Hi dhopp97,

First of all you have to disable Spanning Tree on MLT ports cause using both options is not recommended and interfere with aggregation setup. Also what have you been monitoring which went through only one of the mlt ports? Based on the source and destination mac-address, it can happen multiple sessions do take only one port of the mlt (hash algorithm).

Regards,
JeroeNortel
 
It was actually working. I didn't realize that it did a source/destination mac comparison. I thought it did more of a per-packet balancing.

I have read quite a few documents that have said to leave spanning tree enabled with MLT (one of them I believe was from
--Dennis
 
I was a little fuzzy on that as well, and seem to remember my local support engineer saying that it should be enabled.

Having said that, I talked to him again yesterday and he pointed me to the latest Converged Campus Technical Solution Guide is fairly explicate about shutting off STP on the uplinks of edge switches whatever the configuration - see sections 2.2.3.1.1 and 2.2.4.2.1

cscat=DOCDETAIL&DocumentOID=353196&RenditionID=REND340215

So based on that I'd say either my memory is bad (highly unlikely <smile>) or they've changed their tune on this topic, has anyone else had similar conversations recently?
 
I have done several tests including STG enabled on (S)MLT's (fiber). As far as i see there has been no problem with this setup.
However as far as we know and were advised from our supplier and Nortel Networks STG has to be disabled on agregation links because there could be interference in this feature. STG could mark a mlt as a loop hence disabling al except one port in this agregation. I'm not sure if this is still the problem in newer releases though still STG should not be used when using MLT's. Besides, why running STG when you have MLT :)

Regards,
JeroNortel
 
Well we are going to create a physical ring with 4 switches. All have a 2 Mbit MLT to 2 other switches. So without STG we will have a loop problem. But the way I understand it is STG should properly disable one of the MLTs to stop the loop.

Please correct me if I'm wrong.

--Dennis
 
In a case where you're about to create a ring of layer two switches you'll want spanning tree so that one of the links gets shut down. In this configuration you'll want to understand which switch is going to be your root bridge so the links that are least-used get shutdown under normal operations.

I was talking about cases where your layer 2 switches are hooked together in a configuration where they star out from a central switch. In that case those uplink interfaces should not have spanning tree enabled, but the end-user ports should still have spanning tree enabled. The idea is that if a loop is created out in "user land" you want those ports to shutdown instead of your uplink ports.
 
FYI All:
MLT is NOT, I repeat NOT a load balancer. Multi-link trunking, and other proprietary port bundling schemes (like Fast Etherchannel from you-know-who) do indeed use a binary calculation of source/destination MAC or IP dependent on if the operation is Layer 2 (bridging) or Layer 3 (routing).

It is more of a load sharer which will help with multiple traffic flows. The important point to remember is that for a particular flow, the path will always be the same and hence only one of the links in the MLT will be used, regardless of how many links in the MLT one has.

LACP support enhances the MLT abilities but is not available on the code you're at. I'd recommend a later code release as 3.22 has no LACP support.

I hope this is helpful.
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top