DEGXA-TA Gigabit Card

[David Gutierrez_4]

Hello everyone, Could someone tell me if there is there any known issues with the following Gigabit card made by Broadcom?

bcm0: DEGXA-TA Gigabit Ethernet, hardware address: 00-08-02-91-4E-50
emx2 at pci1 slot 2
FCA-2354 : Driver Rev 2.14 : F/W Rev 3.81A4(1.01A0)

[Ralf Puchner]

OS Type and version?
which machine?
emx2 is not Gigabit Ethernet, so what is your intension?

Have you checked if card is supported by your machine (supported option list), which firmware and slot is required for the OS-version (HP webpage) and machine? Have you checked if latest patchkit is installed?

[David Gutierrez_4]

We have the latest software and hardware revisions.
5.1Bpk4
Firmware revision: 6.9-2
PALcode: UNIX version 1.92-33
bcm0: Link up via auto-negotiation (1000 Mbps, full duplex)
bcm0: DEGXA-TA Gigabit Ethernet.

# sysconfig -q bcm
bcm:
PCI_Option = PCI_SE_Rev - 0x210, Vendor_Id - 0x14e4, Device_Id - 0x16ae, Driver_Name - bcm
DriverVersion = V1.0.23 NUMA lanlog
bcm0 at pci2 slot 2

The environment we have is a ES45 Basically my question is trying to find out if any one else has a NFS Cluster with two ES45 using these model of Gigabit card as their primary interface. We seem to be experiencing some issues concerning the performance.

[Eric van Dijken]

Are you also using the "bcm" cards as cluster inter-connect? Or do you use Memorychannel as an interconnect medium.

If one was to ask me to build a NFS cluster, i would use:
2 Gigabit networkcard (bcm) with netrain, per node.
MemoryChannel interconnect, as thats 70% faster than GB.

Performance would also degrade when the "owner" of the filesystem the NFS share is on, is not the active NFS server. File access would than first pass the cluster interconnect. Which you would notice performance wise.

[David Gutierrez_4]

Memory channel II is used as the ics cluster interconnect.
I will look into re-enabling netrain.

David

[Johan Brusche · Accepted solution]

Netrain won't do any good for the network performance since the 2nd gigabit would stay in standby.
LAG (link aggregation) on the other hand would help to increase the LAN troughput.

NFS server traffic can only be focussed towards the node that serves the filesystem via the usage of a secondary cluster alias joined only by that node.

__ Johan.

[David Gutierrez_4]

Johan,
Thank you for the reply. Definately, lag is what will look into for network trunking.

David