Gb NIC only runs at 100Mb

[Luqman Achmat]

Hi all hp-ux/network fundis.

I have 3 x Itanium rx2600 servers, all 3 running hp-uxv3 11.31 (primarily for Data Protector).

1 of the servers only has the option of running its Gigabit Network Interface at 10 or 100Mb. I am unable to set the Gigabit NIC to run at 1GB.

The other 2 servers run fine at 1Gb. Urgent assistance would be much appreciated!

Some output:

/root>ioscan -funC lan
Class I H/W Path Driver S/W State H/W Type Description
====================================================================
lan 0 0/0/3/0 intl100 CLAIMED INTERFACE Intel PCI Pro 10/100Tx Server Adapter
lan 1 0/1/2/0 igelan CLAIMED INTERFACE HP PCI 1000Base-T Core
/root>

/root>lanadmin -s -x 0
Speed = 100000000
Current config = 100 Full-Duplex AUTONEG

/root>lanadmin -s -x 1
Speed = 100000000
Speed = 100 Full-Duplex.
Autonegotiation = On.

/root>

Please assist

Best regards
Luqman

[Steven E. Protter]

Shalom,

The NIC cards are configured correctly.

I would think one of the following is true:
1) It is plugged into a 100 BaseT switch
2) The switch port on a gigabit switch is configured for 100 BaseT

SEP

[Luqman Achmat]

Hiya

Thanks for your rpompt response!

Nope, that is not the case. The switch is a Gb switch and when I set the port to 1000Mb (1Gb), connectivity to the Itanium server breaks.

I suspect a driver problem in hp-ux, because in "smh" (previously called "sam" in hp-ux v1 & v2), there is the option to set to 1000Mb for the other 2 servers. The highest this problematic server can be set to is 100Mb.

It just so happens I am the network admin and have more experience in networks than hp-ux = not a network/switch problem.

Regards
Luqman

[Ramesh S]

Hi

You could check your

--> switch supports 1Gb speed?

--> Try set the speed manually.

lanadmin -S

--> check the speed setted on file /etc/rc.config.d/hpigelanconf

Thanks & Regards

Ramesh

[Pete Randall]

Both the switch and the NIC should be set to auto-negotiate - neither should be forced to anything.


Pete

[rick jones]

Assuming all three systems are in the same place and connected to the same switch...

First, take-back that hardcoding of the speed on the switch - leave auto-sense on - that is the default for HP-UX NICs and auto-sense and autoneg are supposed to be solid these days.

Next, swap the cables between a system that gets 1g and the system that is at 100BT. Do this at the system end and see if 100BT stays with the cable, (implying something amis with the cable or the switch port) or with the system (implying something amis with the NIC).

[Luqman Achmat]

Hi Rick

Thanks for the response. I tried swapping the cables, but the results are the same, the machine still runs at 100Mb, even after another reboot.

I also tried setting the ports on each side statically to 1GB, but then the machine drops off the LAN and never recovers, not until I put everything (or at least the server's NIC) back to auto-sense.

I still suspect the driver, kernel/os. Please have a look at the 2 files in the zip attachment.

Regards
Luqman

[Steve Lewis]

Your ioscan is getting a description of the product straight off the card using a hardware PCI bus probe.

"Intel PCI Pro 10/100Tx Server Adapter"

You only have a 100mbit card installed in this slot. You won't be able to make it go at 1gig.

It has taken the intel 100mbit driver as it should.

You can also use the lanscan command to list all interfaces. Check that the card listed above is the one that you are trying to configure. Maybe there is a 3rd card installed in the server in another slot which is 1gig. You may simply be trying to configure the wrong card.

[Luqman Achmat]

Hi Steve

It is the other card I am trying to configure = "igelan" lan1 (not lan0):

/root>ioscan -funC lan
Class I H/W Path Driver S/W State H/W Type Description
====================================================================
lan 0 0/0/3/0 intl100 CLAIMED INTERFACE Intel PCI Pro 10/100Tx Server Adapter

lan 1 0/1/2/0 igelan CLAIMED INTERFACE HP PCI 1000Base-T Core


Help anyone!!??

Luqman

[Steve Lewis]

Please post the output of lanscan.

[Luqman Achmat]

Hiya - lanscan output as requested:

/root>lanscan
Hardware Station Crd Hdw Net-Interface NM MAC HP-DLPI DLPI
Path Address In# State NamePPA ID Type Support Mjr#
0/0/3/0 0x00306EF38B2F 0 UP lan0 snap0 1 ETHER Yes 119
0/1/2/0 0x00306EF36E10 1 UP lan1 snap1 2 ETHER Yes 119
VLAN5006 0x00306EF36E10 5006 UP lan5006 snap5006 10 ETHER Yes 119
VLAN5005 0x00306EF36E10 5005 UP lan5005 snap5005 9 ETHER Yes 119
VLAN5004 0x00306EF36E10 5004 UP lan5004 snap5004 8 ETHER Yes 119
VLAN5003 0x00306EF36E10 5003 UP lan5003 snap5003 7 ETHER Yes 119
VLAN5002 0x00306EF36E10 5002 UP lan5002 snap5002 6 ETHER Yes 119
VLAN5001 0x00306EF36E10 5001 UP lan5001 snap5001 5 ETHER Yes 119
VLAN5000 0x00306EF36E10 5000 UP lan5000 snap5000 4 ETHER Yes 119
/root>

[Steven E. Protter]

Here are the choices.

1) Card is broken
2) You misconfigured it. Or another admin.
3) My original answer which you decided was valueless.

Leave this one unassigned until you find out I was right.

;-)

SEP

[Luqman Achmat]

Thanks for your input Steven. My money is on the misconfiguration.

I've logged a call with HP for assistance.

Or if anyone has any other ideas, I'm open to them.

Regards
Luqman

p.s. Everybody's input is valuable! :)

[rick jones · Accepted solution]

Occam's razor would suggest that perhaps the cables to the 100BT and the 1G are switched :) I suppose that could be confimed by disconnecting the interface you believe to be the 1G and checking via lanadmin that the operational state goes to down (operational, not administrative state).

It will be very interesting to hear about the final resolution after the official support folks are done.

[Andrew Rutter]

hi there,

this sounds very similar to a support call i had to deal with a while ago. went through all the obvious things and reconfiguring and even down to patches installed and running a lan script from hp. I wasnt on site as it was 600miles round trip and everything seemed ok, apart from not coming up at 1gb. we even had hp's back line support looking at it and couldnt find anything wrong and had info from cisco switches and everything analized. the guys on site were sure everything was right with the network and it was the server config fault. got them to swap the cables and still the same? but this servers other card was like yours 100mb only. eventually tied it down to the some of the cables through the building were only cat5 not cat5e.

if you have another server with 1gb card try and link the 2 together and see if it comes up at 1gb to confirm the card. if it does.. replace the cable with an upgraded one.

Andy

[Luqman Achmat]

Rick/Andy

Solution found! Your suggestions solved the problem.

MachineA = problem server running at 100Mb
MachineB = server running at 1Gb

Though I had swapped the cables between the 2 machines, I still had the same problem on MachineA. However, I decided to plug in a brand new CAT5e cable into MachineA and rebooted. MachineA came up with a 1Gb connection! Strange that MachineB runs at 1Gb with the "faulty" cable from MachineA, though I did not try a reboot on MachineB when swapping the cables. Only rebooted MachineA.

Thanks to all for your help! It is very much appreciated!

Regards
Luqman

[Luqman Achmat]

changed cable to new CAT5e cable and now running at 1Gb.

Luqman

[rick jones]

Down at the physical layer, with these speeds over unshielded twisted pair, it is my understanding that it starts to become rather, well, mystical. If something is close to the margins, it doesn't take all that much to push it the rest of the way. Knuth only knows what things will be like with 10G over UTP :)

I'm not sure how much they cost these days, but you might want to look into getting a cable tester, and go through your stock of cables (spares and those installed) and check them and get rid of the ones which are marginal. It will probably save you some headaches later-on.