I currently have a client that is configuring a new network entirly via CMS from a stack(3) of 3750's as the cluster commander. The 3750 stack is the concentrator for a group of 2950 switches. Each 2950 switch has both of its gig uplink ports coming back to different stack members. The aim is to ultimately configure cross stack etherchannel on the stack to the spoke switches, for redundancy. This has been tested and seemed to have worked fine.
When the client deploys the spoke switches to their final location and connects both legs back to the stack, the following occours:
All links connect with green LED's on all switches. (Keep in mind that etherchannel is not yet configured, and he has applied the "switch macro" via CMS to all uplink ports.) After around 10-15 mins one link LED will turn orange for a few mins and then go out. Around 2-5 mins later the second uplink turns orange and then goes completely out. During all of this CMS performance is horrible. The only fix is to cycle the switches. The network is currently configured with a single leg to each spoke and is working fine. But, I want etherchannel in, as I have informed him it can be done.
As etherchannel is not yet configured, I think it may be something to do with STP perhaps. The switch macro adds "spanning-tree link-type point-to-point" to the config, which is only recommended for use with rapid-PVST+ implementations. These swithes use PVST+ by default, and the macro doesn't seem to change this. Could this be the issue?
I intend to remove the macro's from the uplink interfaces and configure cross stack etherchannel, one spoke at a time, to see if it stablizes. Has anyone experienced this before?
3750 SMI (12.2(18)SE1)
2950 SMI (12.1(20)EA1)
Understanding of this issue would be great.
