Etherchannel - Redundant 6500 to 3750 datacenter switch

[joe.daguiar]

I have 2 6504's running HSRP as my core.  They are each etherchannel'd to my Datacenter switch (3750 Stack) -- see image below.

The problem i a having is with the etherchannel status:

Core1 PO11 status w (waiting)

Core2 PO11 status P(bundled)

DC11 PO48 status P -- but only to Core2 - the interfaces to Core1 are suspended. (See attached configuration documents)  None of the devices have any information in the logs.  I run this same configuration in my central location, but we are running Nexus 7000's.  With the 6500's, do I need to split the port channels on the 3750 to allow them to negotiate the etherchannel? If I split the portchannels, are there any concerns? Should I expect to see the etherchannel status as P (Bundled) or H (Hot-Standby)?

[glen.grant]

  No you do not need to split the port channels for them to work although a better design would be to run one to each switch so that if something happened to the switch or uplink you would still have a path back to the core.  You would probably have to post the relevant configs for all the interfaces involved so that we could see . Both the port channel SVI and the ports themselves.  Just sounds like a misconfig .  Are these port channels a single vlan or are they trunked?  Are they forced on or using lacp ?  Now that I look at it you need to create a different port channel number on the 3750 for one of your links they cannot both be port channel 48 . Create say port channel 49 and put 2 links in that.  You cannot have the same port channel going to 2 different switches.

[Kyle McKay]

Glen is correct.... Your design is as if you had a vPC or VSS in your core - which I don't believe you have.

You would need two completely seperate port-channel's on the 3750 in order to make this work - one to each of your core switches. Your current configuration has one logical link from the 3750 connecting to 2 completely seperate devices.. This is a no-no unless VSS or vPC is implemented.