Hi,

I have some linux systems sitting behind a Cisco CSS load balancer. Every 10 seconds, the CSS connects to an application specific port on the linux systems to verify the "service" is up.

I've noticed that over time, these services crash or stop responding. This appears to be because of the sockets not being recovered after the CSS disconnects from its check. A TCP dump reveals that the CSS is doing a three way tcp handshake on the port, and then immediate sending a reset packet and disconnecting.

Now my question is, how can I determine if this is a bug in the APPLICATION or in the 2.4.21 tcp stack? And has anyone heard about something similar in this or related kernel versions?

Thanks!