Cisco 3750-E Stack Design

rfranzke · ‎12-09-2009

Would like to see if anyone would know if the following would work when using the 3750-E series switches in stack. Currently we use dual 4507R switches in a collapsed core configuration. We have access layer switches which uplink to each 4507 for redundancy. The 4507Rs are uplinked to form the redundant core network. 4507-A fails and all traffic uses 4507-B. Pretty common so far. I would like to know if you can use a 3750-E stack in place of the 4507R. You would have two 3750 stacks which would be uplinked together via one of the switches in each stack. Then access switches would uplink to a switch in each stack. Would this work like I think? I understand that this is not really needed because you can run ether-channel between multiple switches in the stack for the access layer uplinks which would do essentially the same thing but I would like the stacks to be in separate cabinets for power redundancy. Not sure the stack cables would be long enough to span racks. Really just want to know not so much if this is a good design but rather if it would even work with the dual stacks replacing the 4507Rs. Thanks in advance

vvasisth · ‎12-09-2009

I am not sure if i got you correctly but cross stack ether channel will do what you are trying to achieve redundancy.

However there is a catch even if we have cisco devices on both ends we cant use PAGP we have to use LACp for cross stack ether channel.

In any case you would need atleast 12.2(25)SEC or higher inorder to configure a
cross-stack etherchannel. Here are the links for the same:

Release Notes:
http://www.cisco.com/univercd/cc/td/doc/product/lan/cat3750/12225sec/ol7454.htm#wp731959

- PAgP does not support cross-stack EtherChannels.
- After grouping the links into an EtherChannel, LACP adds the group to the spanning tree as a single switch port. 
- You should use care when using the 'on' mode. This is a manual configuration, and ports on both ends of the EtherChannel must have the same configuration. If the group is
misconfigured, packet loss or spanning-tree loops can occur.
- If cross-stack EtherChannel is configured and the switch stack partitions, loops and forwarding misbehaviors can occur.

This little information is an excerpt from the link below where you can find all the information regarding configuring Etherchannels in your 3750:

Configuring EtherChannels and Link-State Tracking
http://cco-sj-2.cisco.com/univercd/cc/td/doc/product/lan/cat3750/12225see/scg/swethchl.htm#wp1275628

rfranzke · ‎12-09-2009

Thanks for the reply. I might not have made myself clear here.

Normally we have two switches and hosts (or switches) connect to each switch either by NIC teaming(host) or using RPVST+ STP(switch). The switches we use happen to be single modular devices cross connected together to form our network core. If I were to replace these modular devices with a stack instead and then connected the two stacks together to mimic the way we do things now, would they work the same as using the modular switches? The stacks are supposed to operate as one switch. I understand the Etherchannel solution would essentially get me the same thing for redundancy but my servers do not support Etherchannel. Etherchannel setup would not be 'cross-stack' but instead use a single stack assuming by cross-stack you mean two stacks connected together and running Etherchannel to one switch in each stack.

rfranzke · ‎06-15-2010

Maybe this is a better question. Is it better to use the stacking capability of the 3750-E series switches, and run cross-stack Etherchannel between your access switches and two 3750-E series switches which or stacked or run spanning tree and uplink your access switches to each non-stacked 3750 switches? Seems that spanning tree, despite cisco's speed extensions to the original protocol, is becoming more and more a thing of the past and the new way to achieve switching redundancy at the hardware level is to make your core switches appear as one switch (by stacking/VSS) to the access switches and then Etherchannel across the VSS/stacked core. Thoughts?

sstrohmayer · ‎07-25-2010

If I understand the post correctly, we are asking how to best configure and cable up a stack of access-layer 3750's to two stacked 3750's acting as a collapsed core. Surely there has to be some best practice for this type of requirement somewhere. I have looked and cant find anything. The problem I have is keeping the network as simple as possible whilst still maintaining redundancy and resiliency. Ideally a full layer-3 environment would be best for me but I simply wont have the skillset available to manage the environment after handover to the customer.