1000Base-T card will not autoneg to 1000G

[Daryl Rose]

I'm not sure of the model number of the card, its a two port 1000Base-T card that will not auto negotiate to 1Gig. The switch is set to autoneg, but the best this card will autoneg to is 100 half. I even tried to force it to 1G, full, but that wouldn't work.

Any suggestions?

Daryl

[Hein van den Heuvel]

Bad cable or port?
Try an other one, which is known to have successfully negotiated up to gigabit speed.

What system? RXxxxx ?
What OS / version? HPUX?
What switch?
How did you verify ? (software, lights on adapter, lights on switch, query to switch )

Hein.

[Patrick Wallek]

You should also verify that you are using a Cat 5e or Cat 6 cable.

Are you also positive that the switch supports 1Gb?

[Daryl Rose]

Hello, and thanks for the quick reply.

Well, I doubt that its the cable, port or switch. For starters, there are actually two identical servers with the same issue. Both are INTEGRITY RX4640 servers.

I inherited these from someone else, and originally they were set to 100/half. I ran new cable myself to a known good switch, but the best I can get is 100/full.

The servers are both HP-UX 11.23. I'm new to HP-UX, so I'm not really familiar with a lot of the commands. I've tried making all of the adjustments using the sam command. If there are other command line commands that will work better, please let me know.

Thanks.

Daryl

[Mel Burslan]

Known good switch does not guarantee you anything other than the label saying it a gig-e capable one. What you need is to veify with the network administrator that the port is set to 1000 AutoNeg. Also as Patrick pointed, cabling is very important for gig-e connections. Considering you are in the radio frequency ranges, interference from external signals are a real big concern, especially if you have a hunkin-big generator or welding machine or something similar running nearby. So, don't trust the cat-5 cable you purchased from best buy or the network cabling left behind from the old days of networking when 100BT was known to be the blazing fast connection.

[Daryl Rose]

I just physically verified that the cables are both cat 6. Brand new, I took them out of the package myself, and no they are not cheap best buy cables. These are high quality cables bought from a known quality cable source. (btw, they are only about 12' in length.)

As for the known gig switch; yes, I've been working with my network admin on this issue. This switch has other systems plugged into it, and they are all running at 1gig, and there are no large pieces of equipment that could be causing this issue.

Both the switch and the server were set to autoneg and the best the servers would do was 100/half, so I took them off autoneg and locked them down to 100/full.

Someone mentioned the MTU setting. Currently the MTU is set to 1500. Is there a better rate that I should try?

Thanks

Daryl

[Duncan Edmonstone]

So first off, show us that these are GbE capabli NICs by giving the output of:

ioscan -funC lan

HTH

Duncan

[Daryl Rose]

Class I H/W Path Driver S/W State H/W Type Description
========================================================================
lan 0 0/1/2/0 iether CLAIMED INTERFACE HP AB352-60001 PCI/PCI-X 1000Base-T Dual-port Core

[Hein van den Heuvel]

>> 'm new to HP-UX, so I'm not really familiar with a lot of the commands

Check out the lanadmin command.

Hein.

[Daryl Rose]

Check out the lanadmin command.

Hein.

Not really sure what you want me to look at here. I don't really see anything that helps, but here is the output of the display command:

Enter command: display lan0

LAN INTERFACE STATUS DISPLAY
Thu, Sep 10,2009 09:18:49

PPA Number = 0
Description = lan0 HP PCI 1000Base-T Release PHNE_36236 B.11.23.0706.02
Type (value) = ethernet-csmacd(6)
MTU Size = 1500
Speed = 100000000
Station Address = 0x306e5d951e
Administration Status (value) = up(1)
Operation Status (value) = up(1)
Last Change = 31136
Inbound Octets = 3671001963
Inbound Unicast Packets = 26282522
Inbound Non-Unicast Packets = 30479
Inbound Discards = 0
Inbound Errors = 0
Inbound Unknown Protocols = 652
Outbound Octets = 3229165162
Outbound Unicast Packets = 19461502
Outbound Non-Unicast Packets = 372
Outbound Discards = 0
Outbound Errors = 0
Outbound Queue Length = 0
Specific = 655367

Press to continue

Ethernet-like Statistics Group

Index = 1
Alignment Errors = 0
FCS Errors = 0
Single Collision Frames = 0
Multiple Collision Frames = 0
Deferred Transmissions = 0
Late Collisions = 0
Excessive Collisions = 0
Internal MAC Transmit Errors = 0
Carrier Sense Errors = 0
Frames Too Long = 0
Internal MAC Receive Errors = 0


Daryl

[Mel Burslan]

Since you mentioned using SAM to configure interfaces, there is a chance that SAM might have mangled the configuration. COuld you please port the contents of this file :

cat /etc/rc.config.d/hpietherconf

[Daryl Rose]

Per your request:

############################################################################
# The HP_IETHER_INIT_ARGS are reserved by HP. They are NOT user changable.

HP_IETHER_INIT_ARGS="HP_IETHER_STATION_ADDRESS HP_IETHER_SPEED HP_IETHER_MTU HP_IETHER_FLOW_CONTROL HP_IETHER_AUTONEG HP_IETHER_SEND_COAL_TICKS HP_IETHER_RECV_COAL_TICKS HP_IETHER_SEND_MAX_BUFS HP_IETHER_RECV_MAX_BUFS HP_IETHER_VMTU HP_IETHER_ITR_MODE"

HP_IETHER_INTERFACE_NAME[0]=lan0
HP_IETHER_STATION_ADDRESS[0]=
HP_IETHER_SPEED[0]=100fd
HP_IETHER_MTU[0]=1500
HP_IETHER_FLOW_CONTROL[0]=1
HP_IETHER_AUTONEG[0]=0
HP_IETHER_SEND_COAL_TICKS[0]=150
HP_IETHER_RECV_COAL_TICKS[0]=0
HP_IETHER_SEND_MAX_BUFS[0]=1
HP_IETHER_RECV_MAX_BUFS[0]=1
HP_IETHER_VMTU[0]=0
HP_IETHER_ITR_MODE[0]=-1

HP_IETHER_INTERFACE_NAME[1]=lan1
HP_IETHER_STATION_ADDRESS[1]=0x00306E5D951F
HP_IETHER_SPEED[1]=100fd
HP_IETHER_MTU[1]=1500
HP_IETHER_FLOW_CONTROL[1]=1
HP_IETHER_AUTONEG[1]=0
HP_IETHER_SEND_COAL_TICKS[1]=150
HP_IETHER_RECV_COAL_TICKS[1]=0
HP_IETHER_SEND_MAX_BUFS[1]=1
HP_IETHER_RECV_MAX_BUFS[1]=1
HP_IETHER_VMTU[1]=0
HP_IETHER_ITR_MODE[1]=-1

# End of hpietherconf configuration file

[Matti_Kurkela]

HP_IETHER_INTERFACE_NAME[0]=lan0
HP_IETHER_SPEED[0]=100fd
HP_IETHER_AUTONEG[0]=0

HP_IETHER_INTERFACE_NAME[1]=lan1
HP_IETHER_SPEED[1]=100fd
HP_IETHER_AUTONEG[1]=0

According to this file, both lan0 and lan1 seem to be explicitly configured to 100 Full-Duplex, autonegotiation disabled.

Your lanadmin command output also confirms this (straight copy-paste):
Speed = 100000000
Adding some spaces to make it more obvious:
Speed = 100 000 000
i.e. 100 Mbps.

With this configuration, if the switch side is set to auto-negotiate, it will detect the 100 Mbps link pulse but gets no response to autonegotiation. In this situation, the standard says the switch must pick 100-Half.

MK

[Daryl Rose]

Matti said:
"According to this file, both lan0 and lan1 seem to be explicitly configured to 100 Full-Duplex, autonegotiation disabled. "

Your are correct. The current settings are 100 full. As I stated earlier, the original settings were 100/half autoneg off. I re-cabled to a 1gig switch and changed autoneg to on. The network remained at 100/half. I tried to force it to 1g/full, but received a message telling me that would not work. The best I can get at the moment is 100/full with autoneg turn off.

Thanks.

Daryl

[Bill Hassell]

100-half is the fallback value defined by industry standards when autonegotiate fails. Since you have a short cable (less than 100 ft), then the switch is highly suspect as to setting or function. If you have a small Gbit (dumb) switch, plug the cable into that switch, then use lanadmin -X auto_on where is the lan number (ie, lan0 or lan6) like this:

lanadmin -X auto_on 6

Then read the resultant negotiation after a minute or so:

lanadmin -x 6

(note: -X to set, -x to read results)

Note also that 100 HD with a switch that is set to 100 FD will result in massive error counts and about 1/20 normal throughput. Use this command:

lanadmin -g 6

To read the current stats.

[Daryl Rose]

Well, I really have a hard time believing that this is the switch, but I'll get with my lan admin and see what we can do.

Thanks for the help.

Daryl

[Andrew Rutter]

hi,

the quickest and easiest way to rule out the switch, is just connect the two together with a crossover cable.

set the system to Autoneg, and then see what they come up at.

connected together if they come up at 1gig, then its a problem beyond the unix.

if they dont then I would start looking at the patches possibly for the cards

Andy

[rick jones]

If you hardcode 100fd in the file, you are implicitly disabling autonegotiation at the HP-UX end. This will mean that the switch's attempt to autonegotiate will fail and it will fall-back to 100hd (or at least should).