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

hung network

Jan-Arendt Klingel — Tue, 28 Mar 2006 01:43:59 GMT

I've 2 L-class servers connected to a switch, then to a router/gateway. Recently, access to one server was lost. But the other server was fine. After re-booting the router, access to the server was fine again. My question is if it is possible for only the path/route to that server to get hung? Such that the only solution is to power cycle the router?

Re: hung network

Muthukumar_5 — Tue, 28 Mar 2006 01:54:07 GMT

check the network usage between connection lost server and router. Is there any error messages? Physical connectivity may be too?

--
Muthu

Re: hung network

RAC_1 — Tue, 28 Mar 2006 01:55:58 GMT

Unless, you find out the root cause, you should not be doing that. What exactly happened? From server side, check what happened. Check network logs. Anything in syslog.log?
netfmt -f /var/adm/nettl.LOGxx

Re: hung network

Ninad_1 — Tue, 28 Mar 2006 04:42:09 GMT

Hi Jan,

Check few things
1. Any messages on console or /var/adm/messages or syslog files on the server for which connection is lost
2. When connection to one server is lost do you mean that you are not in the same network as the server is ? what does able to reach the router ? do you check through ping ? or how do you check.
3. Are you able to reach other servers in the same network ?

Ninad

Re: hung network

Kent Ostby — Tue, 28 Mar 2006 14:39:13 GMT

It could happen if you've got a bad cable or possibly even a bad card on the l-class.

If the hardware is only "a little broke" then a reset might be enough to get things going again.

I'd check/tighten the cables and make sure the board is seated for that machine.

Re: hung network

rick jones — Tue, 28 Mar 2006 20:56:50 GMT

Unless ip_ire_gw_probe has been cleared via ndd, HP-UX 11 at least used to "probe" the configured gateways to make sure they were still "there." This probe was an ICMP echo request (aka ping). If the gateway was configured to ignore ICMP echo requests, the HP-UX system would end-up thinking the gateway was dead and disable that route.

Perhaps someone changed ndd settings in nddconf on one of the systems, or perhaps they have different patch levels.