HP ProLiant Onboard Broadcom NIC negotiates 100mbps only

[CakeDestroyer]

Hi guys,

I have been having some strange link speed problems with some of the ProLiant servers. It seemed to have started after some recent firmware updates but I can't pin-point them exactly.

Basically on some of my ESXi 5 hosts, the onboard Broadcom gigabit NIC only negotiates 100 mbps link speed. The Intel Pro 1000 NIC in the same server work fine. I have tried different cables, switches etc but same thing happens. I have noticed this problem on the following servers:

1 x ProLiant ML150 G6 (NetXtreme BCM5723)

2 x ProLiant ML115 G5 (NetXtreme BCM5722)

I have just used HP Firmware Maintenance DVD 10.10 to update firmware and the problem stays.

Has anyone else encountered this problem? Is there anything within ESXi I can do to test/investigate this problem?Any tips/advice would be greatly appreciated.

thanks

[MartinAmaro]

Have you look at "Configuring the speed and duplex of an ESX/ESXi host network adapter"

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1004089

[CakeDestroyer]

Hi,

I had another look at the documet you linked and although it didn't provide a fix it did help me find some more info.

Both of my switch and ESXi host NIC are in auto negotiation mode.

In Networking -> Properties on vSphere client, I could manually set the NIC to use 1000 full (or half) duplex. However after I set this, the link shows up down. If I change it back to 100 full, the link shows as up.

I also logged in to the management GUI on my switch, the switch port says:

Neighbour advertised capabilities: 10 Half 10 Full 100 Half 100 Full

Very strange.

[BharatR]

Hi,

Please share the Output of

esxcfg-nics -l

[CakeDestroyer]

Here we go:

~ # esxcfg-nics -l

Name    PCI           Driver      Link Speed     Duplex MAC Address       MTU    Description

vmnic0  0000:01:04.00 e1000       Up   1000Mbps  Full   00:1b:21:4f:48:7f 1500   Intel Corporation 82541PI Gigabit Ethernet Controller

vmnic1  0000:04:00.00 tg3         Up   100Mbps   Full   f4:ce:46:fb:87:f8 1500   Broadcom Corporation NetXtreme BCM5723 Gigabit Ethernet

[BharatR]

Hi,

vmnic1  0000:04:00.00  tg3         Up   100Mbps   Full   f4:ce:46:fb:87:f8 1500   Broadcom  Corporation NetXtreme BCM5723 Gigabit Ethernet

This is an Add-On Card ? can u share me the HP Model no on the card.

[CakeDestroyer]

Hi,

No this is no an additional card. This is built-in to the motherboard. The model of the server on which I ran the command is HP ProLiant ML150 G6. The broadcom is the integrated single port gigabit NIC.

Regards

[BharatR]

Hi,

Try to update the BIOS and Ethernet Adapter firmware in the latest version for this Server and the Driver.

to check the driver version the command is below.

ethtool -i vmnic0

driver:  bnx2

version:  2.0.7d-4vmw

firmware-version: 5.2.3 NCSI  2.0.6

bus-info:  0000:02:00.0

Perhaps the physical switch is blocking the host from communicating (such as  with Cisco BPDU guard)? You could try plugging in the ethernet cable into  another physical switch to ensure it's an issue on the ESX host, or check the  switch config to make sure no blocking is occuring or can set manually at 1000mbps for testing.

[MartinAmaro]

I'm sure what BharatR sugguested is the best way to resolve your issue. Can you also find out what is the peed duplex configured in the swtich port?

That way you can make sure you match the speed duplex config of the vmnic1.

[CakeDestroyer]

Hi guys,

I just booted into HP Maintenance DVD 10.10 again, which is dated 4 June 2012. All the BIOS, NIC are all running the latest version of the firmware.

The only thing other thing I thought of is maybe there's a newer version of the firmware, which is not integrated to the DVD yet. However the version numbers on the hp web site doenst seem to match up with the ones on the DVD. Also the firmware updates on HP's web site is on-line updates... which is only available on Windows or Linux. And obviously I am running ESXi.

I could find another drive to install Windows just to try the firmware update. But I rather not do that, especially I seem to remember the last time I tried it the firmware wasn't actually meant for my particular NIC.

Also, the switch port was set to auto-negotiate. I have tried other switches too, unmanaged switches. Same behaviour...

Thanks for your ideas so far though.

Regards

[kastlunger]

Hi;

I am experiencing the same issue on a Proliant dl160 G5.

NIC will set itself to Autonegotiate 10 Mbit half, despite it being connected to 1000Mbit switch.

Also, changing the setting to any value shuts the NIC down.

Here the relevant info :

~ # ethtool -i vmnic0
driver: tg3
version: 3.110h.v50.4

firmware-version: 5722-v3.07, ASFIPMI v6.02
bus-info: 0000:03:00.0

~ # esxcfg-nics -l
Name    PCI           Driver      Link Speed     Duplex MAC Address       MTU    Description
vmnic0  0000:03:00.00 tg3         Down 0Mbps     Half   00:1e:0b:dc:1c:c0 1500   Broadcom Corporation NetXtreme BCM5722 Gigabit Ethernet
vmnic1  0000:04:00.00 tg3         Up   10Mbps    Half   00:1e:0b:dc:1c:c1 1500   Broadcom Corporation NetXtreme BCM5722 Gigabit Ethernet

Also, I happened to find something similar from HP support that might be related (although the driver version mentioned is older):

http://bizsupport1.austin.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=us&taskId=110&prodSeriesId=3659898&prodTypeId=329290&objectID=c01886839&printver=true

[stacycarter]

Surprised I did not see this last year. Did you ever find a resolution? I saw this same type of behavior on many Proliant DL380s which happened to be running ESXi 4.1 U3 at the time.  Seen across different clusters, different switches, and the problem was intermittent (did not happen with every auto-negotiation). Only had the problem w/ the integrated 1Gb Broadcom nics, the Intel nics In these servers never had problems.  The model of nic and driver were different in my case (5709 based, bnx2 driver).  Went through all the same troubleshooting steps: made sure auto/auto was configured, updated firmware/drivers, swapped cables.  Unfortunately, the vendors could not reproduce, and concluded "it's not our problem". And when I would mention it could be a driver problem, I would get told "no no no that's not possible, this is at the hardware level".  If that's the case, why do I see old KBs for similar issues where a driver update is the fix?  Still a mystery, though not much of a concern at the present time.

[Big_Banana]

Sad that this issue has languished on this site so long.  I too am "afflicted".  I have had my ML150 G6 for a while.  VMware Server 1.0 and 2.0 worked great, NIC was running a 1Gbps under Windows 2003 and later 2008.

I decided to go to ESXi 5.0 and debugging slow file copies I find that the NIC is locked to 100Mbps.  Changing it manually (vShpere or Console) turns the card off.  Switch lights even go dead.

Definitely a a driver/OS problem, since this only happened after changing OS's.  (I did use the HP branded ESXi ISO)