I've got a question re: the ACE module.
Lets say I have 2 web rservers and I have a probe interval for checking them from the ACE of 10 seconds.
Lets say a probe just passed and it is 10 seconds before the next one. The ACE will think the rserver is ok. Then say the rserver httpd service is stopped at 3 seconds after the last successful probe, therefore leaving 7 seconds before the ACE is going to send another probe. The ACE will think it is still 'up' before the next probe is sent.
Given the above, what happens to a) existing connections to the newly failed rserver and b) new connections if the failure occurs between probes?
How does the ACE handle this situation?
Are there any differences between how the ACE handles this between A1 and A2 versions of software?
Existing connections will be unaffected by what happens to a rserver by default.
You can configure a 'failaction' under the serverfarm to force a disconnect or a reassign.
New connections will still be forwarded to a server that has not yet been detected as down.
Ace will just send the SYN.
Your client will retransmit and ACE will again forward the SYN.
This will go on until the client gives up or the server is detected as dead and the connection will then by forwarded to a different rserver.