Strange ping response

[Kamal_12]

Hi everybody.
It happens a very strange behavior to my alphaserver with Tru64 5.1A.
I just reinstalled the operating system with the minimum software possible.
In whatever mode i set the NICs (with console command "set ewa0_mode $mode") they respond in this way (pinged from a linux box, the same from a m$ box):

[root@beta /]# ping alpha
PING alpha (192.168.0.250) from 192.168.0.130 : 56(84) bytes of data.
64 bytes from alpha (192.168.0.250): icmp_seq=2 ttl=64 time=0.222 ms
64 bytes from alpha (192.168.0.250): icmp_seq=4 ttl=64 time=0.216 ms
64 bytes from alpha (192.168.0.250): icmp_seq=6 ttl=64 time=0.208 ms
64 bytes from alpha (192.168.0.250): icmp_seq=8 ttl=64 time=0.205 ms
64 bytes from alpha (192.168.0.250): icmp_seq=10 ttl=64 time=0.196 ms
64 bytes from alpha (192.168.0.250): icmp_seq=12 ttl=64 time=0.189 ms
64 bytes from alpha (192.168.0.250): icmp_seq=14 ttl=64 time=0.188 ms
64 bytes from alpha (192.168.0.250): icmp_seq=16 ttl=64 time=0.181 ms

--- alpha ping statistics ---
16 packets transmitted, 8 received, 50% loss, time 15012ms
rtt min/avg/max/mdev = 0.181/0.200/0.222/0.020 ms

As you can see, one time it answers, one time not. It gives me problem on a ftp connection too. I tried many times this test, always the same answer.
Could it e a hardware problem? (Strange: the same behavior on two NICs).

What's happening?

Thank you

[Michael Schulte zur Sur]

Hi,

can you post ifconfig -a?
Did you by accident assign the same ip to both NICs?

Michael

[Johan Brusche]

This is typical for a hdx/fdx mismatch between system and lan-switch. So check with
hwmgr -get attribute -category network, what the OS thinks it is set to. If it is different from the switch-port setting, try the lan_config command to find the correct setting, eg /usr/sbin/ln_config -i ee2 -a 0

If your devices are of the type "eex", and your switch is autonegociate-capable, then leave the console ewa0_mode = auto.
If the lan-devices are of the "tux"-type then use fixed settings at both ends.

Rgds,
Johan.

[Ravi_8]

Hi,

There is no problem in h/w

could u post the o/p
#ifconfig -a

[Ralf Puchner]

johan,

the ee devices doesn't honor any console switches. The intel driver is purely autoneg oriented!

kamal:
please post the output of "netstat -i" and name the used device. It seems 100% a network problem (connection settings or ip related, like duplicate IP's).

[Kamal_12]

Thank you. Here is the config.

Kamal

[Michael Schulte zur Sur]

Hi,

was that your intention to have both cards in the same subnet?
See also:
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=440321

Michael

[Kamal_12]

Thank you, but now one card is not connected. The problem I told is the same on both cards and I don't use them simultaneously.

Kamal

[Michael Schulte zur Sur · Accepted solution]

Hi,

if one card is not connected, what does it do with a packet, it can not deliver?
I do not know for sure, how Tru Unix handles two cards for one subnet, if it does round robbin. But try to deconfigure the nic, that is not linked and try the ping again.

Michael

[Ralf Puchner]

michael,

ping works on MAC base, so if one adapter is not connected there is no sideeffect

[Kamal_12]

I can try... How can I deconfig a NIC?

Tnx

Kamal

[Michael Schulte zur Sur]

Hi,

use sysman -> networking -> network setup -> deconfigure

deconfigure the nic, you want to stop.

greetings,

Michael

[Kamal_12]

Deconfigured the NIC. I don't know why, but it works!!!

Thank you very much.

Kamal

[Ninad_1]

Its strange ,
I have an ES40 server I have also configured 2 NIC cards with ip addresses in same subnet and only one is active at a time but I dont face any kind of problem
I wonder what exactly solved the problem

Ninad

[Hein van den Heuvel]

Kamal> Thank you, but now one card is not connected.

Ralf> ping works on MAC base, so if one adapter is not connected there is no sideeffect

Yeah but the MAC address only comes in play after the drivers selected a target NIC no? It woudl seem that the OS selected the non-connected nic every other time.

Kamal> Deconfigured the NIC. I don't know why, but it works!!!

Great. Thanks for the feedback.

Ninad> I have also configured 2 NIC cards with ip addresses in same subnet and only one is active at a time but I dont face any kind of problem

But in your case those nics are properly connetced to the real network. Not so for Kamal.

[Ralf Puchner]

Hein,

it seems he is using an OS 5.x because in this version latent support is enabled. This feature allow some kind of "load balancing" without any failover mechanism (so a problem leads to network outage). In v4.x latent support is disable by default. If using latent support switches are necessary and routing daemons must be disabled (see network admin guide)

But why not using LAG? It is more robust, true load balancing.

In case of ping an ICMP will be automatically answered by low-level functions, I'm just wondering why the answer of the interface will be spreaded over all interfaces....it makes no sense.