1000FD - not able to require with auto_on

[DuKul]

Hello all,

I have two Combo adapters:

"HP AB465-60001 PCI/PCI-X 1000Base-T 2-port 2Gb FC/2-port 1000B-T Combo Adapter"

and one embeded card "HP PCI 1000Base-T Core"

on 9000/800/rp3440 HP server.

The cabling is done to Cisco Catalyst switch.
On Catalyst there is speed/duplex setting auto/auto.

However when setting the speed of the lan port to auto_on (Autonegotiation = ON) only the embeded card is behaving well - it gets 1000FD speed.
The remaining two lan interfaces are not able to acquire the correct speed i.e. 1000FD.

This could be an issue with old cables (not reliable with 1000fd...) or with the LAN switch, but I am not sure.

[Steven E. Protter]

Shalom,

Replacing the cables is always a good idea.

Ideas:

1) Perhaps have the cables moved to another switch port.
2) Check with networking that the ios, the os on the router is updated.
3) Re-affirm that the network port is auto/auto and the router is capable of 1000FD
4) HP-UX is was until recently incapable of having two interfaces up on the lan on the same network. If your OS is unpatched you may be experiencing this issue and be able to rectify with the isntallation of a current bi-annual patch set.

SEP

[Tommy_6]

Before you go wasting a lot of time on this, check the cables. I had a similiar problem two weeks ago. I ended up changing the cable and everything starting to work. Let us know if that fixes the problem.

[Steve Lewis]

cat 5e, not just cat 5, for best reliability.

[rick jones]

HP-UX has always allowed someone to configure multiple physical NICs into the same IP subnet. It just has not/does not always do what one might initially expect. And that is only affecting "things IP" it would have no bearing on whether or not a given NIC would autosense to 100 or 1000 Mbit/s.

[DuKul]

Thanks guys for your answers still. I am in contact with the person who is responsible for the Cisco Catalyst side of the issue :-)
It seems that there is an autonegotiation set on the switch, but only on certain speeds/duplexes.

Various NICs are not a problem at all - HP-UX is pretty good at this :-) It even provides wonderfull APA which is the target also for me once having this strange issue solved.

[rick jones]

Perhaps it is their popularity, but the name Cisco seems to pop-up with considerable frequency in discussions of duplex/autoneg issues.

[DuKul]

Unbelievable :-(

I have made a network cisco guy to set the ports in various ways - i.e. autonegotiation for any speed, autonegotiation for exact speed/duplex, strict speed/duplex setting... but to no avail. Only one of the interfaces works well - the embeded one which does not come from Combo card, but this could be just coincidence, but who is there to tell?
Moreover when setting the ports in various ways I have been getting various values for duplex when autonegotiating on the side of the server.

I will check the cables......

[DuKul]

Unbelievable :-(

I have made a network cisco guy to set the ports in various ways - i.e. autonegotiation for any speed, autonegotiation for exact speed/duplex, strict speed/duplex setting... but to no avail. Only one of the interfaces works well - the embeded one which does not come from Combo card. This could be just coincidence, but who is there to tell?
Moreover when setting the ports in various ways I have been getting various values for duplex when autonegotiating on the side of the server.

I will check the cables......

[rick jones]

Don't forget that if you hardcode one side, you _have_ to hardcode the other side the same way, you cannot leave autoneg on on one side and hardcode the other. So, when you had the network admin set the switch to 100FD you have to set the NIC to 100FD too.

[DuKul]

exactly - this is the final setup - to set 100FD manually on both sides.

[rick jones]

I thought you wanted 1000FD? Were the cables or the switch just not up to the task?

[DuKul]

Ye I need 1000FD but for the meantime we have to fix the connections to an stable and acceptable values. This is because this server is a part of MC/SG cluster with very important package running on it.

[DuKul]

so I will do what Tommy recommended - to check the cables...

[DuKul]

Hi all,

gosh, this took lot of time :-(
But finally, yes after several months of arguing, the third party had checked the cables and it was there - the problem :-).

So after re-cabling everything is as it should be.

Dukul

[Pete Randall]

I'm glad you solved your problem and I think it is truly wonderful that you came back to the thread after four months to update the conclusion.

Now, if you would just assign points to the folks that gave their time to try to help you. Tommy, in particular deserves a nice fat bunny.


Pete

[Bill Hassell]

And for everyone wondering about cat5, cat5e and cat6...

http://www.connectworld.net/syscon/support.htm

I would rip out *ALL*cat5 cables when upgrading to Gigabit speeds. Some short ones may work for a while but like all transmission lines, accuracy at both ends critically depends on length, total capacitance, and crosstalk. These may be digital bits but they are being transmitted using analog methods. An oscilloscope or TDR scope would have shown the problem cables quite easily.