this week i have tested the deployment of a N1kv on our UCS test system.
Everything worked fine after i have disabled the fabric failover option of each vNIC in the service profiles.
With the failover option enabled the VEMs could not connect to the VSM.
In this white paper ( http://www.cisco.com/en/US/prod/collateral/switches/ps9441/ps9902/white_paper_c11-558242.pdf ) it is recommended to enable the failover option. have i missed something important - any hints from your side?
ESX4 - latest and greatest
n1kv: version 1(2)
VSMs as VMs on ESX Servers on UCS-Blades
thx for your suggestions!
There should be no issue enabling failover on your vNIC within UCS. All this does is create a "Stanby" veth on the other Fabric Interconnect to expedite failover. I have this setup in my lab with no issues. Can you post your VSM config please.
There have been many enhancements to failover performance in UCS 1.1.1 release, so if this is a lab environment you may wish to upgrade to the latest code before digging too deep on this. (Version 1.1.1l is being released today...)
Like I said I had failover working fine at version 1.0.1e in my setup.
i'll gonna test it once again with my release (1.1(1j) not 1.1.1e - sorry for the mistake) and post the results.
probably i'll have time to test it tomorrow. if there's enough time i'll test the new UCS firmware too.
It is recommended to disable the "fabric failover" on UCS. If you enable fabric failover on UCS, esx will not see the link as down.
This was noticed recently and corrected. The updated doc will be published soon..
hi Robert, Naren,
i had time to test the config once again and experienced no troubles enabling the hw failover on my menlo adapters.
must have been an weird circumstance in my environment....
But im happy to hear that an updated paper will be online (will you also post it in the community?)