LACP Port Channel to 1000v fails after working fine

[l600671]

I am looking to see if anyone else has experianced this problem rather than looking for a solution at this time.

We have a standard design with 2 N5Ks set up with vPC using 2232s.  Hosts connect to the 2232s with dual 10 Gb adapters.  These are VMWare hosts running a 1000v for the virtual switch.  Everything will run fine for quite a while and then intermittently one of many hosts may loose network connectivity and the only indication we get is that the port channel on the N5K has gone from an up state (P) to an individual (I) state.  To correct this problem we reload the driver.  Our belief is that we need to upgrade the driver and the firmware for the NIC but have not found a bug which indicates this exact problem.  Cisco can not find any issue with the 1000v and VMWare can not find any issue with ESX.  The aggressiveness of any driver/firmware upgrade depends on knowing for sure if the problem is fixed because the effort is huge for this volume of servers.

Has anyone else seen a similar issue?

Thanks

[mwronkow]

Hello N1k User,

What version is N1k running?

Could you please list the model & driver version on the host? 

esxcfg-nics -l

ethtool -i

Is this a rack or blade server? If blade, is a passthrough switch involved?

Has a TAC case previously been opened for this issue?  Please send me the number in a private message.

Thank you,

Matthew