server doen't communicate with network

[Bakang Moatlhodi]

I have just upgraded a k380 server from unix 10.20 to hp-ux 11.0. I now have a problem because it is failing to commnunicate with network. i have set the ip address, gateway and the subnetmask. The lan card shows that it is enabled. The product number for the lan card is A3495-60001 .

I need help very ugently..

[Mark Greene_1 · Accepted solution]

What does ifconfig lan0 return, and what is the link light on the card doing?

[twang]

Try to verify LAN connectivity and see:

# linkloop 0x001083E00F50
Link connectivity to LAN station: 0x001083E00F50
-- OK

[Sridhar Bhaskarla]

Hi,

A3495A is HP-PB 10/100 lancard. Check the speed settings on the card.

#lanadmin -x P

where P is the PPA of the lancard. If it is lan1, the ppa will be 1.
If it is supposed to be 100 Full duplex, then change the speed

#lanadmin -X 100FD P

Edit /etc/rc.config.d/hpbasetconf and add the interface and speed as below to survive the reboot

HP_BASET_INTERFACE_NAME[0]=lanP
HP_BASET_SPEED[0]=100FD

-Sri

[Massimo Bianchi]

Hi,
was the correct card plugged ?

Did the card have auto-negotiation ON (which is the default)?

My doubt is that your upgrade phase did a little mess with lan naming, and now it is confiugred a lan that you don'tuse.

Is there any UNKNOW/UNCLAIMED/NO_HW in the ioscan ?

HTH,
Massimo


BTW i'm trying to send this answe for the 3rd time... hope it is not getting here too many times..

[Steven E. Protter]

After Shridar' test, you may find you need to hard code network configuration in the following file:

/etc/rc.config.d/hpgsc100conf

or

hpfcsc_lanconf

depending on NIC type.

Example:

HP_GSC100_INERFACE_NAME[0]=lan0
HP_GSC100_STATION_ADDRESS[0]=
HP_GSC1--_SPEED[0]=100FD


SEP

[Bakang Moatlhodi]

Thank you for your response
Mark
The we are using is lan4 and the output from the command you sent is ok and the link light is just on. What is it supposed to be doing?

twang
linkloop command you said I should try is failing

Sridhar
The speed of the card is 10

Massimo
The the card is plugged in correctly.
When i do an ioscan it shows that it is claimed ans there is no no_hw.

Steven
I haven't tried you suggestion.

Thank you all, i will be grateful if you send in more answers.

[Christopher McCray_1]

Hello,

I noticed you said there is no_hw; How about insf -e?

Hope this helps

Chris

[Sridhar Bhaskarla]

Hi Bakang,

Just to get it clarified. Were you using 10 HalfDuplex or 100 Full Duplex before on 10.20?. If it was 100 Full, then you can use the command

lanadmin -X 100FD 4

to reset the card to 100 Full Duplex.

-Sri

[Mark Greene_1]

If the link light is on solid, you have a good physical connection. If it is blinking, you have a bad connection. If it is out, you have a bad card. It sounds like you have a good physical connection, and your problem is most likely configuration.

Verify that the switch/router is set to the same speed and duplex as the card.

mark

[Bakang Moatlhodi]

we have always been using 10. (it was working fine all along)

[Massimo Bianchi]

Hi,
check also with a

arp -a


just to see if there is something cached, and you get nowhere due to some network mismatch.

HTH,
Massimo

[Sridhar Bhaskarla]

Hi Bakang,

In that case what does your "lanadmin -x 4" report?.

For 10 HD, AUTO Negotiation off will not work as far as I remember. Try enabling AUTO_ON and let the card detect the switch setting.

#lanadmin -X 4 AUTO_ON

Reset the card

#lanadmin

->lan-> ppa 4 -> reset

-Sri

[Dario_1]

Hi!

What's the output of ioscan -fnC lan?

Thx,

Dario

[Ron Kinner]

The usual cause of your problem is that the default gateway you are trying to use will not respond to a ping. 11.0 thinks it must be dead so it won't use it. To turn it off as a test:

ndd -set /dev/ip ip_ire_gw_prob 0

If it won't take it then you need a patch.

To make sure the change stays after a reboot edit /etc/rc.config.d/nddconf to add:

TRANSPORT_NAME[0]=ip
NDD_NAME[0]=ip_ire_gw_prob
NDD_VALUE[0]=0

If you already have entries then use the next available integer in the brackets.

Another common problem with 11.0 is the sudden enforcement of an obsolete RFC which said you couldn't use subnet zero. Subnet zero is the bottom subnet in a division of a large subnet. Example: I take a 192.168.2.0 255.255.255.0 subnet and I divide it into smaller pieces by using a 255.255.255.252 subnet. The first subnet would normally be 192.168.2.0 - 192.168.2.3 but the old RFC said this was not legal. The first legal subnet you could use would be 192.168.2.4 -192.168.2.7.

To fix:

ndd -set /dev/ip ip_check_subnet_addr 0

and also modify /etc/rc.config.d/nddconf:

TRANSPORT_NAME[0]=ip
NDD_NAME[0]=ip_check_subnet_addr
NDD_VALUE[0]=0

Change the number in the bracket to the next higher number if you already have entries in nddconf.

Finally, get the MAC of the gateway and run a linkloop test.

linkloop -i ppa MACofGateway

Where ppa is the ppa of the NIC that connects to the gateway (if you have more than one NIC if only one then you just need linkloop MACofgateway). This will test your layer two connectivity. If this passes then it's a layer three (IP) issue. Check that you have the correct IP address and Mask! Also disconnect your NIC and have the gateway ping the IP address of your box. IF this works someone is squatting on your IP address.

If the linkloop fails then it's a NIC/driver/cable (are you plugged into the correct NIC - often if there are more than one then UNIX's idea of how they are numbered may be different than yours.)/Switch-port problem. Do
lanadmin
lan
display
to see if you have two "UP"'s. (use ppa x to change to the correct NIC if you have more than one).

Ron

(second try. first one gave a 404 error. I went back and looked and there was still no sign of my post so am trying a second time.)