BladeSystem - General
1748171 Members
4583 Online
108758 Solutions
New Discussion

Re: Virtual Connect Stacking + Onboard Administrator cabling.

 
chuckk281
Trusted Contributor

Virtual Connect Stacking + Onboard Administrator cabling.

Soon Heng was looking into how the various stacking methods are used for VC and the OA.

 

**************************************************************************************************************

If we stacked 4 VC together.

 

The VC mgmt traffic does not used the stack cable but the OA link.

 

Which one do we support:

  1. Using OA stacking, up and downlink to stack all VC together
  2. Having all enclosure OA connect to an external switch and used that switch and used it for the VC mgmt traffic.

******************************************************************************************************************

Dave wanted to clarify some things about stacking.

***************************************************************************************************************

Soon Heng,

                Stacking the VC Ethernet modules and stacking the OA service links are two different and independent items.

 

                Stacking the VC modules allows internal ‘Production’ traffic to traverse within the 2, 3 or 4 stacked enclosures, reducing the need/use of external switch ‘uplink’ ports.

 

Stacking the OA service ports can reduce the number of logins to the OA modules, as it will use the service ports to login/manage the enclosures. You can connect up to seven enclosure with this method. You can use this whether the enclosures employ VC, pass-through or industry switches. These ports use APIPA addresses(169.254.x.x/16), so you would not connect this port to an external switch. You would not be able to login to a remote enclosure’s iLO through these ports as they are on the non-routable APIPA address.

 

It would be your customers choice to use service port stacking(easy login/configuration) or the OA(access to iLO) link to an external switch. Both are supported.

**********************************************************************************************************

 

Does this help? Clear up some things about stacking? Do you have other questions?

2 REPLIES 2
chuckk281
Trusted Contributor

Re: Virtual Connect Stacking + Onboard Administrator cabling.

Additional conversation regarding this topic.

 

From Cullen:

 

********************************************************************************************************************

 

Dave,

 

So to clarify here:  we know that within an enclosure it’s important to create crosslinks between VC Ethernet modules, but that you can manage VC Ethernet modules (and flash them) without crosslinks.  Presumably the management traffic goes across the internal network, not through the crosslinks.

 

When you stack VC modules in different enclosures, you don’t need to also use the OA uplink/downlink ports.  Management traffic from the VCM in enclosure 1 will traverse the stacking links to control the modules in enclosure 2.

 

Is that more or less right?

 

***************************************************************************************************************

 

To which Dave replied:

 

*************************************************************************************************************

 

Cullen,

                When you flash the VC FW, you will do it through each enclosures OA IP address using the VCSU util(As the GUI is going away in VC 3.0). No need to traverse enclosures.

 

I believe that the VC configurations are communicated through the OA NICs(need factory/developers to verify) as the OA stacking links would not pass it(non routable IP) and the VC stacking links are for normal network traffic through the internal server NICs, not management type communications.

 

With the c-Class enclosure we have three different types of external connections:

  1. Interconnect ports  – These would all be used for uplink connections to a customer’s Core/Edge or SAN switches. This is what I would consider ‘Production’ type communications or connections to their VLANs/ SANs.

 

  1. OA NIC ports              - These are the RJ45 NIC ports on the OA. These would be connected to the customers ‘Management’ VLAN / network. As Best Practice we should always attempt to keep ‘Production’ and ‘Management’ traffic separate. No ‘Production’ traffic would go through these ports. There is no connection to the ‘Production’ or interconnect ports.

 

  1. Service link port        - These are the RJ45 ports on the OA tray/sleeve physically between to two OA slots. These will have an APIPA address which needs to be kept SEARATE or not connected to a customer’s network. This is only to link multiple enclosures together for management use only. The 169.254.x.x/16 IP is non routable for connection to the iLO or interconnect IPs. It allows for some management traffic only and allows a laptop to be connected to the ‘top’ enclosure’s link. The laptop link can be used to log into OAs if you have a password.

 

The Service links are optional

 

**************************************************************************************************************

 

Good conversation. Comments?

TurbinYA
Occasional Advisor

Re: Virtual Connect Stacking + Onboard Administrator cabling.

You can find accurate information about this in HP Virtual Connect for c-Class BladeSystem Setup and Installation Guide

 

 

 

The VC Manager in the primary enclosure accesses all remote modules and OAs over the management

 

network. The OAs for each enclosure to be managed as part of a VC domain must be on the same

 

management subnet along with all of the VC-Enet and VC-FC modules that are in each enclosure. All

 

enclosure OAs and VC modules within the same VC domain must be on the same lightly loaded subnet. HP

 

recommends that the OA IP addresses used are configured to be static. The VC-Enet modules use stacking

 

cables between enclosures to route network traffic from any server port to any uplink port within the VC

 

domain.

...

 

When both Primary and Standby modules in the base enclosure are taken down for maintenance or

 

lose power and are no longer present in the domain, the management capabilities in the VC domain

 

are lost. Both the Primary and Standby modules in the base enclosure must be recovered to regain

 

management access to the VC domain.

 

If network and fabric uplinks are defined on the remaining enclosures, the servers continue to have

 

network and storage access.

 

 I think VC can use OA stacking links.