10GbE with 1GbE standby?

[taylorb]

ON paper this should work, but I just wanted to see if anyone has done something like this.     We are upgrading the VM hosts to 10GbE, but the adapters we can find from HP only come in 2 port models.  I would prefer two individual cards to protect against a card hardware failure, but these cards aren't cheap and PCIe slots are at a premium in some of the hosts...   So I was thinking, the hosts all have plenty of 1Gb ports, why not run the dual 10GbE links from the one card and then make a 1Gb port as the standby for the rare occurrence of a total NIC failure.  Worst case, that would keep everything running at slower bandwidth while I have a chance to vmotion and fix the problem with no downtime.  Our storage is all on 8Gb FC, so this would not be storage traffic.   

Sound good?

I'd just add the 1Gb port under "Standby Adapters" and the two 10Gb ports under "Active Adapters" in the NiC teaming tab under the vswitch properties, right?

.

[Josh26]

I would prefer two individual cards to protect against a card hardware failure

How many failed NIC cards have you seen in the last ten years?

I've seen failed RAM, failed motherboards, fan and power supplies, but the last time a NIC failed, it was a 100Mbit model.

You have HA there to deal with a server that experiences a failure - a NIC fail counts as such a failure. I wouldn't let you worry you.

[a_p_]

I'd just add the 1Gb port under "Standby Adapters" and the two 10Gb ports under "Active Adapters" in the NiC teaming tab under the vswitch properties, right?

With the default policy "Route based on originating port ID" this will work fine. The VM itself will always see the same uplink to the vSwitch. What changes in case of a failure is only the uplink to the physical world.

André

PS: With 2 active and 1 stand-by adapter, the stand-by adapter will already be used in case of a single NIC failure!

[taylorb]

Josh26 wrote:

I would prefer two individual cards to protect against a card hardware failure

How many failed NIC cards have you seen in the last ten years?

I've seen failed RAM, failed motherboards, fan and power supplies, but the last time a NIC failed, it was a 100Mbit model.

You have HA there to deal with a server that experiences a failure - a NIC fail counts as such a failure. I wouldn't let you worry you.

Fair point, but for the cost of a 1GB port, why not cover the bases?  HA still results in a short downtime and application crashes.  If I thought it had much chance of occurring, I'd run 2 of the 10Gb cards,. 

[taylorb]

a.p. wrote:

PS: With 2 active and 1 stand-by adapter, the stand-by adapter will already be used in case of a single NIC failure!

Interesting.   So if I drop a switch or bump a cable, the 1Gb link will pop in even if there is still an active adapter available?  That is something to think about.  

Truth be told, like most people, I doubt I'll really be pushing more than 1 gig worth of data very often, so I doubt anyone will notice.