Losing Connectivity

[James Doddery]

Hello all -
Have 3 HP-UX boxes running B11.00. Network connectivity seems to be randomly dropping off. The 3 HP boxes are the only boxes on the switch having this issue, and they only appear
to be having trouble communicating with each other. All are on the same subnet, and when the connectivity drops between the HP-UX servers, all the other servers are still able to communicate with them. Any ideas ?

[Luk Vandenbussche]

Hi James,

Is your networkspeed the same on the switchs and on the server.

I propose to put them on 100 MBit Full duplex fixed.

[James Doddery]

Yes, the Ports on the switch and the LAN cards on the servers are all 100/FD. We tried switching the ports around, to no avail. This has everyone here completely baffled, though I must admin, none of us are UNIX experts.
Thanks

[Charles McCary]

James,

are they off the network completely at this point?

Is your gateway also a firewall? Does it respond to a ping?

thanks,

c

[Chan 007]

Any Recent changes to your systems?

Check your link light..!!

Also check your subnet mask?

Try to do a netstat.

Can you send o/p of lanscan and netstat -nr

Chan

[Bill Hassell]

Talk to your network admins and ask if they decided to turn off ICMP ping response from the routers. It is a common (for security) to turn off the ping resply for routers. Starting at 11.00, HP-UX implemented a dead gateway (router) detection feature that disables the interface when the router doesn't reply to an occasional ping.

To see if this is the case:

ndd -get /dev/ip ip_ire_status | grep -e IRE_GATEWAY -e flag

If you see a dead gateway, then the route has been disabled because the router refuses to reply to a ping. Check the current setting:

ndd -get /dev/ip ip_ire_gw_probe

A value of 1 means that the feature is enabled. You need to add the latest network cumulative patches: PHNE_33395 and PHNE_28525. And to fix a buf in ndd with the -c option, add PHNE_31965. To turn off the feature:

ndd -set /dev/ip ip_ire_gw_probe 0

But to make the change permanent (so it survives a reboot), change the nddconf file (/etc/rc.config.d) to add the setting:

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

If there are already other settings for [0], use the next array reference like [1] or [2]...

[James Doddery]

They don't stay down....they lose the connection, and then anywhere from 30 seconds to 4 minutes, they stay down.

Will try the ndd, and will let you know.

Thanks
-James

[Charles McCary]

Bill beat me to the punch...that's where I was going with the firewall question (generally they've got ping turned off). Anyway, hope that works for you.

c

[Steven E. Protter]

Shalom James,

I like Bill's suggestion. He is right 99% of the time.

Another possibility, less likely is that a machine is going on the network with the same ip addresss of one of your HP boxes.

This would lead to a temporary loss of connectivity to at least the box with the matching ip.

Something to look into.

SEP