03-12-2015 01:09 PM - edited 03-01-2019 12:04 PM
some of the VLANs were deleted from N5K-1 and not -2
About 3 servers stayed up and didn't loose connectivity and some other servers lost connectivity .
the vPC links between the N5ks & UCS stayed up. how UCS can detect such failure in the northbound Ethernet switch?
03-12-2015 02:08 PM
In general, all vlan's defined in a UCS domain, are trunked on all ethernet uplinks. If you delete some of the vlan's on the Northbound switch, UCS's assumption that it can send traffic on any uplink (portchannel) will fail, for those blades / service profiles / vnic's, which are are pinned to the uplinks with missing vlan's.
03-13-2015 06:30 AM
Thanks Walter. that is why i am asking if there is any mechanism exist today to stop UCS from forwarding specific VLAN traffic to northbound switch which it doesn't have that VLAN? or if i have failed VLAN in the northbound Ethernet switch can the UCS be notified not to send traffic for that VLAN?
03-13-2015 10:51 AM
I exclude disjoint vlans for a moment;
All vlans in a UCS domains have to be configured on the Northbound switches, see eg.
http://www.cisco.com/c/en/us/products/collateral/switches/nexus-7000-series-switches/white_paper_c11-623265.html
Traffic between vlan's has to be routed, external of UCS, UCS fabric Interconnect only do L2 !
But there is another case, L2 only, where traffic has to exit UCS domain Northbound: one blade / vnic is connected to fabric A, talking to a blade / vnic which is connected to fabric B, both on same vlan.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide