10-08-2012 03:12 AM
I have an old server that I will use to something and I can't use the integreted network cards, and have bought an NC326m network card.
But have will I get this to work in my C7000 enclosure?
In interconnect bay 1 and 2 I have Cisco Catalyst 3020 switch and in bay 5 and 6 I have HP 4Gb Fibre Channel Pass-Thru.
Is it possible to get the server connected to the Cisco switch with the mezzanine card?
I see that the mezzanine slots are connected to different interconnect bays, but I hope it is a way to use the mezzanine card in my setup?
I can't move the Cisco switches, because then the other server will stop working.
Hopefully someone have a solution on this:)
10-08-2012 07:02 AM
There is no way to route a network interface from a mezzanine card to interconnect bays 1 or 2, since the backplane connections are hardwired. If the mezzanine slots 3 and 4 are not populated, you will also need an appropriate I/O module (either a switch or a Pass-Through module).
You did not identify your blade model... but in general, adding the NC326m card to mezzanine slot 1 and populating the interconnect bay 3 and/or 4 would be the only solution to get an actual separate hardware network connection.
But if you can allow the new connection to share network bandwidth with another NIC on the blade, and at least one of the existing 3020's has a free external network port you could use, it would be possible to use VLAN trunking configuration between the blade and the 3020 to add new connectivity to your blade without needing the NC326m at all.
Step 1: find a VLAN ID number that does not overlap with your existing VLANs, and configure the free external port of the 3020 to that VLAN. Connect the new network cable to the 3020.
Step 2: determine how the 3020's internal port connecting to your blade is currently configured. If it's already in trunk mode (passing VLAN-tagged traffic), just add the VLAN number you chose in step 1 to the list of VLANs this port is member of.
If the internal port is in non-trunk mode, you'll need to switch the port to trunk mode. In this case, you can either configure the existing VLAN as a Native (untagged) VLAN for the port (requiring minimal or no configuration changes in the server OS) or switch the existing VLAN to tagged mode too (might keep the overall configuration easier to understand). The new VLAN will be added in tagged mode in both cases.
Step 3: adjust the server's network configuration to accept the new tagged VLAN(s) on the appropriate NIC. You'll need the VLAN ID you chose in step 1, and if you had to change the existing VLAN to tagged mode, you will now need to configure its VLAN ID too.After this, your OS should have an extra set of network configuration settings for the new VLAN, which should behave pretty much like the respective configuration settings for a new physical NIC would.
10-08-2012 11:21 PM
Hi and thanks for your answer.
I don't think I can do as you write.
The only reason I installed an NC326m Mezzanine card is because a colleague need a server that we can set the network interface to 100 MB full duplex.
It can't be done on the integrated interface (BL465c G1) with the netword interface NC370i
Any solution for that? HP have told me that it isn't supported on the NC370i
10-09-2012 12:24 AM
Does your colleague have some strange software that absolutely needs to see the NIC at 100/FD as indicated by the server OS??
If you have no such requirement, you can set the external port of the 3020 switch to 100/FD. The switch may provide some buffering, but the overall performance should be about the same as with a real 100/FD NIC.
The c-class blade NICs (whether integrated or on mezzanine cards) are basically split down the middle: what is on the "inner half" is either integrated to the blade or on the mezzanine card and controlled by the blade OS; the "outer half" is in the I/O module and controlled by it. The speed/duplex settings fall to the "outer half".
The link between the mezzanine card and the I/O module is proprietary. At least according to Linux NIC drivers, it seems to mimic a gigabit fiberoptic Ethernet link, and the parameters are not changeable. Effectively any NIC speed/duplex info seen by the blade OS in the HP c-class blade architecture should be treated as a compatibility placeholder only: only the I/O module knows the true state and operating mode of the external network interfaces.
If you have a NC326m installed and need to have an external 100/FD network interface, I'm afraid a switch module like the 3020 is the only way, since the gigabit Pass-Through I/O modules (= the simplest possible I/O modules) are gigabit-only: they *cannot* be switched to slower speeds, because they are not configurable at all. They act as a set of very simple media converters between the proprietary pseudo-fibre link within the chassis and the actual gigabit Ethernet cabling outside.
If it is any consolation, you definitely aren't the first one in this community that has been burned by the peculiarities of the c-class blade architecture. It is a good solution if you need 4/8/16/... servers with near-identical hardware and are connecting them to a modern infrastructure (gigabit or better LAN). But the architecture places some pretty important limitations if you need just one blade with a unique I/O configuration or need to interface with a 100M LAN or other older infrastructure.