BladeSystem - General
cancel
Showing results for 
Search instead for 
Did you mean: 

Virtual Connect: 802.1q tunneling to ESX with failover

Brian Weatherhead
Occasional Visitor

Virtual Connect: 802.1q tunneling to ESX with failover

I've read the cookbook and manuals over and over and I don't see how to accomplish this.

Where this all get’s confusing is in the fact that we have four VC modules, and would like to maintain some failover. This section of the “VC user guide\Connecting Virtual Connect Ethernet module Uplinks” (page 48):

Mapping multiple networks to a shared uplink port set:

It is also possible to map multiple VLAN-tagged networks to a set of shared uplink ports. The resulting shared uplink port set allows for the minimum number of cables while still providing for link aggregation and failover.

In the following example, a shared uplink port set is first defined to provide aggregation and failover.

Shared_Uplink_Set_A = {Enclosure1:Bay1:PortX2, Enclosure1:Bay2:PortX2}

To me this means that I can aggregate four ports on VC_1 and four on VC_2 and have failover should one switch or one VC module fail. But we would ideally want to pass the tags to ESX and use ESX 802.1q VLAN tagging to support network assignment. Since this prevents having to statically (physically in the blade center) a VLAN to a blades NIC.

But later on in the manual this appears:

“Network Management” (page 118) :

When the 'Tunnel VLAN Tags' option is selected on the Ethernet Settings (Advanced Settings) screen (on page 115) or the 'Enable VLAN Tunneling' option is selected on the Define Ethernet Network screen (on page 103), server VLAN tagging is only supported on networks with dedicated uplinks, and it cannot be used with shared uplink sets.

This to me means that we cannot tunnel the 802.1q trunk to ESX if we intend to have a failover trunk?

Ideally we would want to have four channels bonded on each of the two VC's, four bonded and active on VC_1 tunneling the 802.1q trunk to the ESX blades. Four bonded and standby on VC_2.

Is this just not possible?