We have the same issue on our Cisco devices since the upgrade to NPM 10.2.1
The redicovery of interfaces was set to 60 minutes, this has been changed to 1440 minutes (once a day). I don't know if you can shut this down.
Did you shut down the topology layer 2 and 3 off only on your Cisco devices, or on all devices. I noticed that on all existing devices this has been activated by default. So the issue could be caused by this setting?
I recommend you to open support ticket for this issue. We have already seen very similar issue and therefore there is big probability that our support will be able to help you.
Please don't forget to put ticket number here.
I shut down the topology layer 2, and layer 3 on the Devices that we were having issue with and that kept those Nexus 5000 devices from spiking the CPU at 100%.
we left the topology active for all other devices as they are not having the same issue..