[defect]: set lifetime for aws-nlb site

[joel-aws]

What type of defect/bug is this?

Unexpected behaviour (obvious or verified by project member)

How can the issue be reproduced?

• Enable the aws-nlb site.
• Open >16 TCP connections to the server for i in {1..17}; do nc 127.0.0.1 8080; done
• You will get Failed inserting TCP socket into parent list. errors and new TCP connections will be blocked until the idle_time of 30s is hit, which is usually the Health Check default time.

I believe that by setting the lifetime in the example aws-nlb file to 2 or so, that will suffice in clearing the stack and not letting hanging health-checks clog it up.

Log output from the FreeRADIUS daemon

Reached maximum lifetime on socket status from client (192.168.65.1, 47195) -> (*, 8080, virtual-server=aws-nlb)
 ... shutting down socket status from client (192.168.65.1, 47195) -> (*, 8080, virtual-server=aws-nlb)
Ready to process requests
 ... new connection request on TCP socket
Failed inserting TCP socket into parent list.

Relevant log output from client utilities

No response

Backtrace from LLDB or GDB

No response

[alandekok]

Yes, there should be a flag there saying "close socket immediately", as there's no reason to keep it around.