How do you check corrupt UDP packets in the private SG network ?

[Sammy_2]

In a 2 node SGeRac cluster, individual ruuning the oracle dbase just shutdown at the same time. (HPUX 11.11). Messages in syslog show problems with communication detected. oracle suggested to check for udp packets.


With all these incomplete headers and checksums (below), is that a problem ? Or how do you check for udp packets ?

The other node in the cluster show 0 values for all.

# netstat -p udp
udp:
38862 incomplete headers
38862 bad checksums
0 socket overflows

[Mel Burslan · Accepted solution]

Sounds like the server where this output came from either has a bad network adapter or the switch port it is connected to is bad. Or a third possibility, the speed/duplex configuration is not matched between the server and the switch port.

Get your network people to assist you in troubleshooting this one. They might need to put a sniffer to see what exactly is happening.

[Tingli]

Very often the half duplex and full duplex are not match on both sides.

[Sammy_2]

Thanks to both. I have setup as auto-neg for switches and it is communicating at 1000meg. I will check with network folks to see if they see anyting on the switch side. The numbers from netstat command are not increasing so HP is saying nettrace wont show anyting.