Cisco Nexus 7K's Rx Pause Frames connected to ESX hosts

[darren nixon]

Hi,

Can anyone please advise.

We have a setup of dual Nexus 7K's connected to 3 ESX hosts and on the Port channels connected to these servers , I,m seeing the RX Pause frames counter increasing daily on all  port channels connected these hosts.

The connectivity between the 7K and ESX's are 4x10G connections , with a pair of 10G's to each 7K and vpC for these Port channels. The ESX's have 1000V installed.

I take it the RX Pause frames,  is the ESX telling the 7k's to slow down, as it cannot process.

Do we need to put in flowcontrol on the N7K's.

[Jeremy Waldrop]

Darren, in my experience I have seen flaky behavior with host side LACP/vPC. A more predictable/reliable setup for me has been to not use switch side port channels/vPC for ESX hosts but instead let VMware/N1KV handle the VM distribution across the 2 uplinks.

For N1KV I usually set this "channel-group auto mode on mac-pinning" on the host uplink port profile.

For VMware distributed switches I normally use the default route based on port ID which is similar to the N1KV mac-pinning mode.

ESX/N1KV do a good job or distributed VMs across the 2 uplinks and do a great job of re-pinning VMs to the other link if there is a failure.