Forum Discussion
NimbostratusRST/ACK sent by F5 loadbalancer to the client
Hello we actually have a problem on a specific VS. Context: my VS is a fastL4 profile with very standard configuration. This VS is called each second. And one time per hour we have an error, a TCP/RST is sent by the F5 to the client closing the actual connection.
What we have done: - creation of a specific autonat pool in order to avoid port famine - verification of TCP idle timeout => value is 300 sec
we are in version 10.2.0 which doesn't allow us to use advanced tcp/rst debug.
I am a little stucked at the moment.
Regards,
2 Replies
What_Lies_Bene1Cirrostratus
Could it be a setting on the server, or perhaps a firewall in the path? Are you 100% sure it's the F5 sending the RST and not something else?
C_Guillaume_187yes, I made some tcpdump capture on the F5 to analyse more precisely the observed behavior. It's clear that the RST come from the F5. I checked health monitor and logs, and the server is always available.
My architechture is the following: (A)IP client <-> (B)IP VS - (C)IP back F5 <-> (D)IP server
I have two scenari of failure: first one connnection is established (at TCP level) and client/server are chatting during 10 sec there is no more traffic on the TCP connection D sends a FIN,ACK to C B relays the FIN, ACK to A A reply with an ACK to B B sends a RST,ACK to A <= this is my problem
second one: connnection is established (at TCP level) and client/server are chatting during 1 sec there is no more traffic on the TCP connection A sends a HTTP POST request to B B sends a RST,ACK to A <= this is my problem
If you want I can send the tcpdump (1Mo each)
Regards,
