HP C7000 - VC and OA Queries

[UselessUser]

Hi,

Got a C7000 and 2 BL460c G6's at the mo. We also have 2 x VC 1/10GB Ethernet modules and 2 x VC FC 4GB modules.

The Ethernet modules are installed into Bays 1 and 2, and the FC's are in Bays 3 and 4.

I have updated all firmwares to the latest available..

First question, we have 1 OA module at the moment. I understand it provides control of the blades and KVM access etc, but I am just wondering what the advantage is of having another?? As far as I know if you remove the OA while the blades are running, it has no effect on them?

Second question I am looking for the HP Virtual Connect Fibre Cookbook as referenced here?

http://vinf.net/2008/04/09/how-does-an-hp-fibre-channel-virtual-connect-module-work/

The other questions relate to SAN connectivty... I fully understand I can do load balancing/redundancy techniques using LACP (Aggregation) for Ethernet, but what about SAN...

The people in this document seem to suggest the same cannot be done for fibre? I take it its because they are talking about the card as a standalone unit and not managed by VC on the Ethernet module??

I ask this because as I am playing around in VC Manager, I can create 2 profiles for SAN, each 1 using all 4 ports on the module, so this surely would create 2 16GB trunks, which I would assign in the server profiles one to each hba (My blades have the 2 port emulex mezzanine)...

Confused... again!

[JKytsi]

Second OA is needed for redundancy. Blades will stay on if you take out the OA, but you can't do anything remotely to servers without the OA (use power button,use ILO, etc..)

HP Virtual Connect Fibre Channel Networking Scenarios Cookbook
http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual/c01702940/c01702940.pdf

Usually in corporate SANs you have reduntant SAN fabrics.
Module A to SAN A and module B to SAN B
You can connect up to 4 links per SAN. So 16Gb of bandwith / SAN. And you can enable dynamic load-balancing in VC to those links.

[JKytsi]

... And do remeber to assign point to answers that were helpful to You.

[Rob_69_1]

Second OA is useful if the first one fails.

Regarding FC connectivity, it's just that the same HBA port cannot be mapped to 2 FC switches at the same time. That's why the second schema on the document you link is marked as wrong.

Multipath drivers onboard of the blade server will take care to failover to the second FC path if the first one fails, so you end up using the second HBA port pointing to the second switch via the second VC module.

[UselessUser]

Thanks for the replies... it is good to confirm the change to allow dynamic mode happened as of a later firmware, which is probably why they mention the static mapping in that link I referenced...

Regarding the setup... if I had 4 x 4GB links on each module and created the 2 16GB trunks, what is the best way of allocating traffic...

Is it best to create two profiles, one called SAN and allocate all of Bay 3 ports to it, and one called SAN backup and allocate all of Bay 4 ports to it.

Then on my hosts allocate HBA 1 to the SAN group and HBA 2 to the SAN backup group. Then configure multipathing on ESX to prefer HBA1?

Or are there better designs involving VMware... (First eight blades prefer HBA1 and second eight prefer HBA2, thus giving potentially 2GB's per host instead of 1GB) etc...

[WFHC-WI]

Hi Lawrence,

Either of the scenarios you provide would be acceptable. It is a matter of judging your bandwidth requirements. One other consideration is using Static mode for these connections; that way you can force traffic across specific ports to specific servers. It may help if certain blades have especially high requirements.
Check pages 77-78 in the document here:
http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual/c01730611/c01730611.pdf

In our environment we go for complete fiber path redundancy by installing a second HBA in each server, then installing modules in bays 3 and 5 (leaving 4 and 6 empty). That does mean we only have one linked port on each adapter, but we can tolerate a failure to a single fabric, interconnect, or HBA.

[UselessUser]

Hi,

Thats a good idea, but is having a single module in those bays like that a supported configuration...

Also unfortunately for VMware I need more NIC's so my second mezzanine card will be a quad port ethernet

(What are you doing for this?)

Also I was reading a guide called:

HP Virtual Connect and VMware Infrastructure 3

And it mentions these phrases:

With HP BladeSystem c-Class and Virtual Connect, up to 12 NICs can be utilized within a single server while still maintaining redundant connections to Fibre Channel

(What options set this up?)

In a traditional environment, this series of redundant connections per server means numerous switch ports to manage. For example, 16 rack mount servers with 12 network interface cards each would require 192 network ports. With HP Virtual Connect, an enclosure of eight full-height servers with 12 NICs and a full complement of Virtual Connect modules becomes a single logical device to configure. This environment can be made completely redundant with as few as 8 cables linked to the first layer of managed switches.

How is this done?

[Steven Clementi]

"Thats a good idea, but is having a single module in those bays like that a supported configuration... "

yes, it is suported. unfortunately that does not work well in a VMware solution since you usually always need more nics.


One thing to keep in mind about the OA is that without the OA in place (and no redundant OA), there is no management control over the fans and other devices.

Without an OA monitoring the health of the internal devices, the fans kick into max spin since they don;t know for themselves if the enclosure is too hot. They just assume the worst.


Stveen