HP Virtual Connect : Mapped or Tunnelled VLANs
By no means is this conclusive, but based upon experience and testing of a set of c-Class chassis using Flex-10 Ethernet Virtual Connect Modules.
No limit on VLAN numbers, no need to define individual VLANs on Virtual Connect.
When tunnelling VLANs you lose the ability to be selective about the VLANs passed to an interface. This is likely to increase the number of required uplinks to a blade chassis; Tunnelling is ALL or NOTHING.
Anything that uses the trunk will have ALL VLAN traffic passed through to the server NIC, therefore VLAN’s must be configured on the servers NIC within the OS.
You can take a single defined VLAN and pass it untagged to a server NIC.
You can also pass some or all VLANs to the server, you can be as specific as we like, mapped is very flexiable.
Limitation on VLAN numbers… 320.
From reading through the HP docs I get the impression that mapped is the preffered option.I’ve configured this on 6 chassis with a variety of different servers from ESXi hosts to Ecommerce Web Servers and Clusters, mapped VLANs gives me the flexibility I need to limit the number of uplinks whilst providing as many connections from the VC modules to the blades.
Regardless of Configuration
Switches should be configured to have trunked interfaces using 802.1Q Trunk and 802.3ad LAG protocols, this will allow grouped uplink sets which are active active.
To verify Virtual Connect has formed an LACP LAG, navigate to Interconnect Bays and select the I/O bay where the uplink ports are linked. Locate the LAG ID column, and that the assigned uplink ports share the same LAG ID.