ARP Table Stays Empty

[Ken Rahaim]

I'm working on an HP D Class 9000 (HPUX 11) that I'm unable to establish connectivity with. This is the 4th of 4 servers on this subnet and all others are working fine. Aside from IP address most everything is the same between the 4 servers and I've triple checked the basics (lan0 is up and running according to ifconfig) to the extent that I'm able to.

The one thing I've noticed is there are no entries in the ARP tables. Even when I try to manually add an entry using "arp -s hostname MACAddress pub" nothing shows up when I do an arp -a.

I suspect a mix up with a MAC Address somewhere (there are 3 lan cards on this box) but I've checked the only 2 things that I know that are relevant: 1) lanscan to confirm I'm using the right MAC Address & 2) /etc/rc.config.d/hpetherconf.

Any help would be appreciated.
--
KenR

[John Payne_2]

You may need to look in /etc/rc.config.d at hpbase100conf, hppci100conf, hpsppci100conf, etc. (with the D-class, my guess is hpgsc100conf.) depending on what type of card you are talking about. I accidently did this once in an image, and populated it out to 4 other machines. (Giving us 5 machines with identical mac addresses.) That gave the network guys fits for a week until they isolated the problem and came looking for me.

The problem is if you enter it in or have SAM touch the card, then make an image and keep the file in the image. (Which is what I am supposing happened, since you say that the machines are identical.)

Good luck

John

[harry d brown jr]

There's no need to play with a mac address on an HP to get it connected to a network, unless you have a switch/router that cares. You say you have three lan cards, is only one of them connected to your network? Is the link light lit?

What does lanscan and lanadmin show?

live free or die
harry

[harry d brown jr]

John,

I ditto that one, but on the other hand I had to build a replacement server and it had to have the previous mac.

live free or die
harry

[Ken Rahaim]

John: I should clear up that they weren't cloned. Just that their network configurations (subnets, gateways, etc) are identical (aside from IPs of course). Also, I checked for anything odd leftover in the /etc/rc.config.d config files and couldn't find anything out of the ordinary.

Harry: I am embarrassed to say, but the link light for the TP 10Mb connection is not on.

Here is what lanscan shows:

H/W Address crd# state net int name ppa nm id type dlpi dlpi mjr#

8/16/6 0x080009fd6288 0 up lan0 snap0 1 ether yes 119

8/8/2/0 0x0060b0f98d43 3 up lan3 snap3 2 ether yes 119

8/8/1/0 0x0060b0f98d42 2 up lan2 snap2 3 ether yes 119

Thank you for your help.
--
KenR

[Ken Rahaim]

Oh yes, also, only one NIC is physically connected to the network. The one that is connected is lan0, TP 10Mb.
--
KenR

[Ron Kinner]

Assuming you have a LINK light on both ends of the circuit. IF not did you accidentally use a crossover port or plug into a uplink on the hub/switch?

Try lanadmin
It will bring up a menu. Choose LAN and Display and it will show you the status of the first NIC. Note the speed/duplex and any errors. Then check the others (ppa 1, display; ppa 2 display).

You say you have two other NICs on the HP. Are you sure that both of them are still working? Can someone on the same subnet still ping them? (Don't bother pinging your own address. That seldom proves anything.) I don't know about HP but we had a bunch of SCO machines with 3 NICs and they tended to shuffle the order of assigning the IP addresses during a reboot. We used to have to start a ping on one subnet and plug in network cables until we got it to work and then moved the ping to a host on the next subnet and tried the remaining two possibilities. This also works if someone accidentally plugs the cables in wrong.

Disconnect your NIC and have another host on the subnet ping your IP address. The ping should fail but if it doesn't then you have an address conflict.

If there is a cisco router around on the net do a sh arp on it and search for the same MAC (preferably after do a broadcast ping (ping A.B.C.255).

Finally don't forget to double-check your mask. You didn't accidentally get carried away with the 255's did you and forget the last 0?

Is it a hub or a switch? Could the port be disabled, in the wrong VLAN or set as a trunk? Did you wait 50 seconds for spanning tree to do its thing?

Ron

[Ken Rahaim]

Ron wrote: Try lanadmin. It will bring up a menu. Choose LAN and Display and it will show you the status of the first NIC.
--
I must've missed this before but for lan0 (the port I have connected to the net & the port that has no link light) shows "administration=up(1)" but shows "operation=down(2)". How do I set operation to equal up(1)?

Both administration and operation show up(1) for the servers that are working.

Again, thanks for all the help.
--
KenR

[James R. Ferguson]

Hi Ken:

Look at your /etc/rc.config.d/netconf file.

'lan0' is not specified there (as an INTERFACE_NAME). Otherwise, you should see its operational status=up via 'lanadmin'.
Another possibility is that lan0 is specified but its INTERFACE_STATE is coded "down".

Remember to use 'lanscan' to see the nmid or (on 11.x) ppa value for the lan you are interested in viewing and then using this number, select the appropriate nmid (ppa) in 'lanadmin' (e.g. lan0 might be ppa=4).

Regards!

...JRF...

[Ken Rahaim]

James,

I just double checked again just to make sure, but my INTERFACE_NAME=lan0 and my INTERFACE_STATE is coded to up. Both are entered correctly.

Unfortunately lanadmin for ppa 0 (which equates to lan0 as per lanscan) still shows adminstration status=up and operation status=down.

Are there any other reasons this might occur? Is there anyway to ensure the other 2 unconnected LAN cards are not interfering in some way?
--
KenR

[Ron Kinner]

I expect it means pretty much what it says. The NIC is programmed but not working. Ethernet uses a keepalive sent every 10 seconds and the NIC is not seeing the keepalive. Either you have a bad/no cable, bad card (have you tried moving the IP address to a different NIC?), or bad switch/hub port (swap with a known good port).

Try plugging the network cable in to the other two NICs and check LANADMIN each time. The one you think is LAN0 might not be the one that HP thinks is LAN0.

Ron

[John Payne_2]

Do you have access to a fluke? Maybe it is time to just 'verify' that you have link all the way through the cable.

Or plug it into on of your other machines...

John

[Steven Sim Kok Leong]

Hi,

Since
1) Your hardware status is up in lanscan
2) Your linklight is off

One test you can perform is to connect this network card to another 10 Mbps card via a cross-over cable. Does the link-light comes on and can you ping (to ensure that it is not a LED problem) the destination system you linked to? If it comes on and you are able to ping the other system, can you check:

1) whether the switch port has been accidentally hardcoded to 100FD or 100HD?

2) whether the straight cable you used is faulty by replacing it with another cable.

If you still cannot achieve network connectivity, I suspect that something might be wrong with your card and you should try getting it replaced.

Hope this helps. Regards.

Steven Sim Kok Leong

[Ron Kinner]

Just realized I have Cisco on the brain. Too much studying for my CCNP tests. I said the card sent out keepalives every 10 seconds. That's what a Cisco router does. HP probably does nothing more than check for the link. In any case I stand by my statement that it's probably the card, cable or port. Until you get a link light on both ends (and lanadmin says it's up up) it's not going to work.

Ron

[Ken Rahaim]

Thanks everyone, for their help on this problem. As most surmised it had far less to do with the ARP table than it did with outside sources. Ultimately, the port to which I connected my 10Mb NIC was hardcoded to 100Mb half-duplex settings. My server defaults to full duplex. Once the port was set to autosensing/10Mb, full duplex my comms came up just fine.
--
KenR