Feature: Configurable retry timings for SMF / UPF heartbeats

[Mathias Gumz]

Between SMF and UPF continuous heartbeat requests are sent to detect availability of the associated UPFs. If these heartbeat requests are not replied in a given time frame, the UPF is considered to be unavailable and thus all the sessions are released.

The timeout value at the moment is hardcoded and rather strict:

ergw/src/ergw_sx_node.erl

Lines 643 to 647 in 09e97cb

 send_heartbeat(#data{dp = #node{ip = IP}}) -> 
     IEs = [#recovery_time_stamp{ 
 	      time = ergw_gsn_lib:seconds_to_sntp_time(gtp_config:get_start_time())}], 
     Req = #pfcp{version = v1, type = heartbeat_request, ie = IEs}, 
     ergw_sx_socket:call(IP, 500, 5, Req, response_cb(heartbeat)). 

5s for the first timeout
0.5 time out for retries
5 retries

Which makes it like 7.5s to deduce the UPF is not reachable.

For some deployment scenarios that is a really tight deadline to meet.

With a configurable values the situation could be adjusted to the concrete deployment scenario.

[Viacheslav Katsuba]

Closed. Implemented in release 2.8.3.