Procurve network problems

[scastro]

We have 2xHP Porcurve Switch 2800 series and 8xDL380 servers with RedHat and bonding configured. All the servers are coonected to each switch, with bonding enabled at the OS. Also, each switch is connected with the other with two cables for fully redundancy. In the servers are some TCP/IP applications develop in C++. So, the problem is that we have some problems in the TCP/IP communication, because the programs disconnect each other randomly, even with traffic between the applications. Is there any configuration in the switch that may cause this? Maybe some optimization to make the system connections more stables?

Also, pinging some of the IPs in the network a "DUP" message is generated like in the following example:
me@buzz:~$ ping 172.10.30.26
PING 172.10.30.26 (172.10.30.26) 56(84) bytes of data.
64 bytes from 172.10.30.26: icmp_seq=1 ttl=254 time=0.194 ms
64 bytes from 172.10.30.26: icmp_seq=1 ttl=254 time=0.213 ms (DUP!)

Regards

[Richard Brodie_1]

You have two interswitch cables: you trunked them or turned spanning tree on to prevent a loop, yes?

[scastro]

Yes, is the first we do, set the spanning tree on.

[Evert Goor]

What type of bonding are you using.
Try active-backup first. PRoblem is that you are bonding over 2 switches. And ALB can bring some problems with that.

[Evert Goor]

take a look here about using bonding and how.


http://www.linux-foundation.org/en/Net:Bonding#Switch_Configuration

[scastro]

Evert,
thank you very much for your advice, it worked fine. Pings experienced 10 second delays with bonding in mode 0. Now it seem to be working fine. Thanks,
Sebastian-