topic Re: network problem in Operating System - HP-UX

network problem

piyut_1 — Tue, 15 Feb 2005 02:51:58 GMT

dear,

why connection in hp machine sometime down??
when we check in catalyst, there are no problems and connect very well and in server too.but we can't connect to this ip (we can connect before).
we just thread for enable and disable this interface but we still can't coneect to this interface.my server is superdome that conosists of 5 domain.this server in ist domain and have 3 ip address.we can't connect one interface.can u help me ,please!!

Re: network problem

Sunil Sharma_1 — Tue, 15 Feb 2005 03:05:56 GMT

What kind of NIC you are using if it is 100MBPS card would set it to 100MBPS Full Duplex in both side.

This is known issue when card or switch any of one is set on auto neg. on.

Sunil

Re: network problem

piyut_1 — Tue, 15 Feb 2005 03:17:20 GMT

NIC type is fastEthernet.

Card Type:Fast Ethernet
Autoneg:ON
Speed (million bps): 100
Duplex Mode:FULL

At catalyst, port set with auto.

Re: network problem

Mobeen_1 — Tue, 15 Feb 2005 03:20:58 GMT

Piyut,
Cannot you set AUTONEGOTIATE off ?

rgds
Mobeen

Re: network problem

piyut_1 — Tue, 15 Feb 2005 03:23:47 GMT

mobeen,

i have set to speed 100 full and autonegotiate is OFF.but still not connect.:-(

Re: network problem

Mobeen_1 — Tue, 15 Feb 2005 03:45:18 GMT

Piyut,
As Suresh mentions, do you know if on the other end if the same settings have been done :)

We have had this issue a no. of times and mostly it was to do with AUTONEG

rgds
Mobeen

Re: network problem

piyut_1 — Tue, 15 Feb 2005 03:55:54 GMT

please,

i have been set to autoneg is off and 100 full of mspeed.in catalyst, we set autoneg is on.what wrong ???

Re: network problem

Ron Kinner — Wed, 16 Feb 2005 11:24:28 GMT

You have to set both ends the same or you will cause a Duplex mismatch. If you switch the HP to 100 FULL and leave the Catalyst at Auto then the Catalyst will set up at 100 HALF. If the Catalyst was not seeing collisions or FC errors or lots of runts then the duplex setting was not a problem.

Is there a router involved? HPUX 11.0 turns on dead gateway detection automatically. If the router does not respond to a ping then it will not try to use the router anymore after 3 minutes.

You need to turn it off with ndd:

You can test it with:

ndd -get /dev/ip ip_ire_gw_probe

if it comes back with 1 then dead gateway detection is on. 0 means it is off. Turn it off with:

ndd -set /dev/ip ip_ire_gw_probe 0

Remove the gateway and readd it using route. Seems that once you turn off dead gateway detection the process that removes a gateway from the dead list stops too. It should work now.

Make it permanent by adding:

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

to /etc/rc.config.d/nddconf (if you already have entires in nddconf then use the next available integer in the brackets.)

Ron

Re: network problem

rick jones — Wed, 16 Feb 2005 11:33:16 GMT

UNLESS you see FCS errors when lanadmin reports the interface in full duplex, or you see _LATE_ collisions when lanadmin reports half duplex, you do NOT have a duplex mismatch. Unless you have a duplex mismatch, do not start going around hardcoding things as the next time you move things around you almost certainly _WILL_ start getting duplex mismatches.

Can you be more specific about the 3 IP addresses you are assigning, and if they are on one or more NICs, and more about what you mean by "can't connect one interface?" Do you mean you cannot get one interface to come to the "UP" state? Do you mean you cannot establish TCP connections through that interface?