network getting disconnected

[ssudhir]

hi everybody,

I have uniq problem, I hv a HP 9000 D class runneing 11.00 on it.
server 3:
the ip of the server 121.0.0.4
subnet is 255.0.0.0
class a ,gate way 121.0.0.200 which is L3 Switch.

I have similar another 2 Dclass( Server1 and server2) server running 10.20 with ip 121.0.0.1 and 121.0.0.2 with same gate way . All the servers are connedted to the network thru hub.

i am accessing the all the server frm a different Class C network of 172.25.1.XXX
subnet 255.255.255.0 gate way 172.25.1.10.

Problem:

I am able to ping that server 3 frm another servers( server1 and server2) all the time and able to ping from different network for about 5 mins after booting server and getting time out after 5 mins.
i am not able to ping even to the gate way even.but at the same time i am able to ping server1 and server 2.

no error was found in any logs.


what would be the problem? or do you need any further details to proceed for the diagnostic?

thanks and regards
S.Sudhir

[Ivone Johanna v. Arrago]

Hi,

Can you tell if your server1 and server2 also have a netmask of 255.0.0.0? If not, it can be there's the problem.

Ivone.

[John Dvorchak]

This sounds like your system is configured for dead gateway detection. I would look at this url as this issue has come up from time to time and this is a good discussion to tell you how to detect it and correct it.

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x95d2a848deccd61190050090279cd0f9,00.html

[ssudhir]

Hi Ivone...

thnx for the reply..

i not having the same problem in server 1 and server2 they r working fine.

the system was working fine since yesterday...

thax
ssudhir

[Ivone Johanna v. Arrago]

Hi,

I was thinking about your problem, and I have no clue on what it can be. Have some suggestions:
1 - Same ip address on the network
2 - Hub port with problems
3 - Switch port problems
4 - Routing tables on gateways 1 and 2 incomplete or specific for server2 and server3.
5 - Wrong mac address in arp table
6 - Same mac address on different machines, should never occur, because they have to be unique world wide.
7 - Defective cabling
8 - FireWall which is dropping packets on the network

One question: Are you able to ping from the server to the class c computer?

Ivone.

[Vincent Stedema]

Hi,

I think John's right. If dead gateway detection hasn't been disabled on server 3, then after a certain amount of time (which is 5 mins exactly, I think) the system will remove all routes through gateways that it can't ping. This "feature" was introduced in HP-UX 11.00 and therefore you don't see the same thing on servers 1 and 2.

Try disabling dead gateway detection on server 3 (when you still have network connectivity of course):

# ndd -set /dev/ip ip_ire_gw_probe 0

If after this the network keeps working, add this configuration to /etc/rc.config.d/nddconf. You'll probably find more info on how to do that in the link that John provided.

HTH

Regards,

Vincent

[Anthony deRito]

Did you try changing the ports you are using on the hub. Check out the speed and duplex setting of your hub ports.... make sure they are what you expect. Perhaps the hub is having problems. Do you have enough switch ports to bypass the hub?

[U.SivaKumar_2]

Hi sudhir senguttuvan ,

Can you ping to the L3 gateway from all the servers ?.

If yes , it is not the problem of Dead gateway detection in HP-UX.

regards,
U.SivaKumar

[mvr]

I would pay closeer attention to the routing table.

Miro

[ssudhir]

Hi all

Thax for the prompt reply.

For the all the case which Ivone has mentioned

my ans:

1. no same ip on the network
2.hub port is fine
3.switch ort is fine
3.routing table is same as the other servers
4. arp entry cannot be wrong 'coz of unique ip and mac adress
6.cable is fine
7. no fire wall are configured in network
8. yes i am able to connect to class c n/w from other servers

[ssudhir]

hi every body

i followed the steps for the dead gateway dedection. i am still facing the same proble.

Just for testing purpose i changed the Ip of the server frm class A to class C .Then i discovered a differnt type of problem.Now i am able to access the (ping) the server cont. for hours. but after some tim i am getting soucde quench error.

but when i chaned the IP basck to the same class A . ealier problem contineud..

Any clue where could be the problem

regrds
ssudhir

[Ron Kinner]

Can you ping the router? If not and you have turned off the dead gateway detection with ndd then I think you are missing a patch which should let you turn off dead gateway detection which is what your original problem most certainly is. (There have been several people who have reported the same problem but have not been able to get ndd to fix it.) The alternative way to fix it is to simply get the L3 switch to respond to pings. By the way the default time is 3 minutes (+ 10 seconds) before it declares the gateway dead. You can change it to a much higher number just to prove that that is the problem:

ndd -set /dev/ip ip_ire_gw_probe_interval 1800000

which should change it to 30 minutes.

There is one more rather clumsy alternative. You can use the PC's own address as the default gateway (just tell it that it has a metric of 0 instead of 1) then it will use proxy arp to get to the outside world. This assumes that the L3 switch does proxy arp. Cisco routers do by default but I've never tried it with a L3 switch.

The source quench is a known bug and is corrected with a patch PHSS_17810 or by just telling ndd not to send it.
ndd -set /dev/ip ip_send_source_quench 0

Finally there is another 11.0 gotcha if you change your 121.0.0.4 to a class C. You run into the old subnet-zero-not-allowed rule from an obsolete RFC which 11.0 likes to enforce. Again ndd can fix it. This time:

ndd -set /dev/ip ip_check_subnet_addr 0

So altogether you need to make

/etc/rc.config.d/nddconf

look like this:

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

TRANSPORT_NAME[1]=ip
NDD_NAME[1]=ip_check_subnet_addr NDD_VALUE[1]=0

TRANSPORT_NAME[2]=ip
NDD_NAME[2]=ip_send_source_quench NDD_VALUE[2]=0

Make the above changes on all of your 11.0 machines and reboot and see if it works better.

If not look for new recommended patches. I would suggest
PHNE 26125, 21835, 21767, 17434

Ron

[Steven E. Protter]

Am I missing something?

You have two machines both thinking they own the entire 121 network.

Sure the IP addresses are different.

I'm curious as to why they were configured that way.

Steve
scratching my head....