Understanding VC configuration

[Erik Hochweller]

Hi,

Im a little confused.

This is my setup.

I have a bl460c with graphics extension slot

I have a 10 gig VC switch

Now when i configure a profile for the blade on the switch it all goes fine, i can select a network that i have created (port 8) and assign it to that blade on the first port, just like picture 1. when i start my machine i get 8 network connections of which 1 is only showing that a cable is plugged in. That connection works fine.

Now i have a second connection that i want to add, it goes to a different switch. Do i assign it on the one that is currently unassigned? I ask because when i do that i dont get, either a second connection showing a cable is connected or a new set of connections for that assigned port.

How do i get a second network into the same blade?

Im confused and sorry if this is very simple...

Thanks,

hochi

[Erik Hochweller]

ok i understand now why its 8... 4 for each NIC. However when i look at the overview even though i have assigned two networks only the first one shows as assigned to port one of the first set of 4... the second set of 4 doesnt get a the network assigned...

why?

[The Brit]

Hi Hochi,
Welcome to the forum. First thing, Is your enclosure a c3000, or a c7000?? The answer to your question will be different in each case.
Some basic information: Based on the information you supplied, I assume that your bl460c is either a G6 or G7, and has 2 Onboard 10Gb (Flex)NICs. Each of your flexNICs is logically split into 4 virtual NICs, so you have 8 NICs available. (They are normally named 1a, 1b, 1c,1d, 2a, 2b, 2c, 2d). The numeric part refers to the "physical" port on the server blade, (note it is a dual port 10Gb 'card'), and the letters refer to the virtual NiCs on that Port.

Now this is where the enclosure type becomes relevent. If it is a c3000, then all (i.e. both) of the physical ports map to Interconnect Bay 1, so all of the virtual NICs map to Bay 1.

If it is a c7000, then physical port 1 is mapped to IC bay 1, and physical port 2 is mapped to IC bay 2. This means that the virtual NICs on Port 1 (i.e. 1a - 1d) are mapped to IC Bay 1, whereas the virtual NICs on Port 2 (2a - 2d) are mapped to IC Bay 2.

Looking at your attachment, I see that on the right of the Ethernet Adapter Connections it indicates "LOM:2-a => Bay 2", which implies that this is a c7000.

From what I can see, all that your need to do is assign an Ethernet Network in the box which says "Network Name". I am not sure if you can assign the same network to multiple ports, but you can try it and see. If the answer is no, then just create a second VNet with the same characteristics but with a different name. (I suggest you might consider creating all of your networks in pairs, i.e. VNet_11_A and VNet_11_B, and then always assign the "A"s to the left hand Interconnect (i.e. odd numbered Ports),and the "B"s to the right side, (keeps everything clear and logical)

Anyway, hope this helps.

Dave.

[Erik Hochweller]

Hi,

Tahnks for the quick reply dave. You are correct its c7000 and G6s. Ok so am i correct when i say that the second nic get in reality assigned to the graphics card which is in bay 2 (the actual blade is in bay 1).

Now i have assigned 3 networks and booted the machine and i get a second network interface which is active.. but now its only 5 gig each...

Isnt there a way to make it have two "real" NICs to get the full 10 gig on each?

And does it mean i actully need to assign it dummy profiles as its goes 1a, 2a, 1b, 2b... and since only 1 is in reality active...

Am i missing something?

Hochi

[The Brit]

Normal way to configure these enclosures (for full redundency and best-practice) is to populate the Interconnect bays in horizontal pairs of identical modules. i.e. in your case 2x10Gb switches in Bays 1 & 2. This would make all of your onboard NICs usable. If you only have 1 10Gb IC module in Bay 1, you will only get the 4 virtual NICs on physical port 1, i.e. 1a - 1d.

I'll be honest, I know nothing about the "...graphics extension slot", (I'm not even a proliant guy) so I cant say anything about how that should be set up. I didn't even know there was such an Interconnect module.

Maybe someone else on the forum can address that. But basically, to get use of the other 4 virtual nics, you need to put a second 10Gb switch in IC Bay 2.

Dave

[Mark Wibaux]

You're getting your bays confused. The Flex-10 modules go in to the "Interconnect" bays on the back of the enclosure (numbered 1 - 8). These are the bays that Dave is talking about.

I see from your image that you are actually using WS460 G6 machines (not BL460). However the "Graphics Expansion Blade" is basically the same as the "HP BladeSystem PCI Expansion Blade" for the servers. This unit goes in the "Server" bays in the front of the enclosure (numbered 1 - 16) and then connects directly to the server beside it via the enclosure midplane. This has nothing to do with the NIC connections.

If you only want 2 x 10Gb NICs at the OS level then in the VC profile for the server only create 2 FlexNICs and either set them to 10Gb of bandwidth or just leave them at Auto. These 2 NICs should end up as LOM1:a and LOM2:a (LOM= LAN on Motherboard).
However this only gives you 10Gb to each of the Flex-10 modules not to the outside world. Efficient usage of both NIC ports really depends on how you have your external links setup.
If you have created a single Shared Uplink Set (SUS) and assigned a port from each Flex-10 module you will see that one port is active and the other port is in Standby mode.
My preference to better utilise both NICs at the server level is to make sure at the VC level I have 2 SUS's defined. Each one has at least one external port (if you have LACP support at the switch level you can join muliple external VC ports together but only on a single module, a LACP trunk cannot span modules). See attached image for a logical image of how the links are setup (NOTE: this one is based on FlexFabric modules so it has both FlexNICS and FlexHBAs defined).

Within each SUS you define the same networks but give them slighly different names (I normally add a -1/-2 or -A/-B to the end of the name). You then assign the -1 network to the first FlexNIC (LOM1:a) and the -2 to the other. At the OS level you can then use a NIC teaming utility to allow the OS to balance the load across both NICs.

[Mark Wibaux]

Hopefully the attachment is there this time.