Over the weekend, we replaced 2 4948E in the core with 2 3850-24XS switches running IOS 16.3.5b. We have a 4x1G port-channel going to one of the core ASAs pushing around 400Mbps (176232 pps) and were dropping enough packets to be production impacting. During our call with Cisco, we changed the hashing to spread out the load a bit, as well as adjusting the buffer with "qos queue-softmax-multiplier 1200". Both helped, but not enough. Any suggestions ? It may be possible to return the 3850s and replace them if there's a better solution to this. We also saw CPU at around 85%, which may or may not be related to the dropped packets - I'm currently leaning towards not being related, since our changes didn't affect CPU at all.
No comments:
Post a Comment