LAN/9000 networking error

[ian_60]

My HP9000 N4000 server running HP_UX 11.0, issued a couple of error messages to the console yesterday:
*** 10/100 mb/s LAN/9000 networking ***
Mon Sep 23 BST 2002 17:17:51.970018 DISASTER subsys: BASE100 LOC:00000
<6016> HPCORE 10/100 BASE-T driver detected a transmit command that timed out in slot (Crd In#) 0.

The event was not recorded in the syslog. Though the message appeared, network connectivity was not lost, but I suspect this may have caused an outage we had a few weeks ago. Today I have disabled auto-negotiation (lanadmin -X 100FD 0), after looking at some similar forum threads.

We have a second LAN nic, which is currently not in use, can I use this as a backup if the primary card fails? I have now connected this to the network and it successfully passes a linkloop test, though I have not configured the card yet as I was concerned about other threads warning about using the same subnet mask. (We are using lan0 - HP PCI 10/100Base-TX core ; not currently in use - lan1 - HP A3738A PCI 10/100Base-TX Ultimate Combo)

Also what setting do I need to change so this records in the syslog.

Thanks in advance,
Ian

[Joaquin Gil de Vergara]

To use second nic as backup you need APA software (you must to order)

Syslog is configured in /etc/syslog.conf

see man pages

good luck!

[Ashwani Kashyap · Accepted solution]

To use your second NIC card as a backup , you have two choices . Either use HPAPA ( HP Auto Port Aggregation ) software or you can use service guard to configure your system as a single node cluster and use this HA software to failover your primary IP's to your backup NIC in case of a failure .

ALl network related errors are logged in /var/adm/nettl.LOG00 (01 etc. ) file .
TO view it use netfmt -f /var/adm/nettl.LOG00 > your_file_name . Later you can vi your_file_name .

[Stephanie Nicholls]

There is another option to APA for redundancy but its not seemless so it depends on your business requirements.

Configure the 2nd NIC as normal. Create a DNS alias that relates to the application, ensure all app users use the alias rather than the DNS Name or ip address. When/if first card fails, update DNS entry to point alias to second IP address.

Cheers

Steph

[Brian M Rawlings]

Ian: presuming you are looking for automatic failover, Ashwani pretty much nailed it. A couple more points about this function:
1> APA is LAN trunking, and only works if the switch supports Cisco-style "Fast Etherchannel" LAN trunking protocol (widely used and available, but not guaranteed). Cisco and HP Procurve switches have this, but not every switch does. Trunking multiplexes packets across 2-4 ethernet channels, with automatic failover for all packets if one channel goes down, and load balancing when all channels are in use. It is ideal for your purposes, but must be running on both ends of the links. Incidentally, the trunked link has only one IP address, making it truely invisible when problems develop with one link.
2> MC/Serviceguard's method is different. You connect up the second LAN, as you have, but do not set it up with an IP address. You then inform MC/SG (in its config scripts) that this LAN (by lan ID) is a failover LAN. The failover LAN sits idle and unused until the primary LAN fails. MC/SG then copies the IP stack used by the primary into the memory reserved for the standby LAN, and through black magic and kernel level deep juju, the standby is immediately started with the IP and MAC of the failed primary LAN.
3> I'm not sure that you can change anything to force all LAN errors of this type to the syslog file. As Ashwani pointed out, it keeps its own log, and it tracks a vast amount of info (potentially). I don't know that I'd want it routed to the normal syslog file.

Hope this helps!
Regards, --bmr

[ian_60]

Thanks for your replies, particularly Ashwani's, I will get hold of the APA software, also useful to know exactly where the network errors go.