One of the slower servers in the cluster keeps on making Pound give a "An
internal server error occurred. Please try again later."
I can replicate this by using a simple server sided script:

To simulate this we made a jsp that just sleeps for 30 seconds and it
returns OK
Thread.sleep(30000); followed by 'OK returned now

I increased the timeout to 65 but and the error stayed the same but appeared
after 1min 15secs rarther than the previous 15secs.

Config is now: (NB. Traffic is passed back to a local HaProxy instance for
cookie insertion so HaProxy could also be causing an issue.)

User    "nobody"
Group    "nobody"
LogLevel    0
Client    65
Server  0

ListenHTTPS
    Address 192.168.1.21
    Port 443
    Cert "/usr/local/etc/server1.pem"
        Service
            BackEnd
                Address 192.168.1.21

                Port 80
            End

        End

End

On Mon, 2007-11-12 at 19:57 +0000, Lists Account wrote:[...]

If HaProxy waits until the full response was received it could explain
what you see (from Pound's point of view it sent a request and waited 65
seconds without a response).[...]