[SR-Users] tcp_set_connection_lifetime(1) does not close connection

Hi, for having a maintenance mode, I need to implement some sort of draining functionality. All our clients register via TLS. So I thought, I just answer the next REGISTER of a client with a 503 and kill the tcp connection afterwards, so they need to open up a new connection to the loadbalancer in front and end up on a different server. This is my code: 510 if msg.is_method("REGISTER"): 1 if KSR.pv.get("$shv(MAINTENANCE_MODE)") == 1: 2 KSR.info("In maintenance mode, rejecting registration") 3 KSR.hdr.append_to_reply("Retry-After: 2\r\n") 4 KSR.sl.sl_send_reply(503, "Draining") 5 if KSR.tcpops.tcp_set_connection_lifetime(1) > 0: 6 KSR.info("Set Connection lifetime to 1") 7 return False (Side note: msg is some internal class, this is obviously KEMI with python) So now when I set the MAINTENANCE_MODE var to 1, Kamailio rejects the REGISTER requests, logs both lines and sends out a 503. Debug log tells me this: Dec 10 15:32:10 ifens5 /usr/sbin/kamailio[3851]: DEBUG: tcpops [tcpops.c:229]: tcpops_set_connection_lifetime(): new connection lifetime for conid=1: 908866403 Dec 10 15:32:21 ifens5 /usr/sbin/kamailio[3855]: DEBUG: <core> [core/tcp_main.c:4554]: tcpconn_main_timeout(): entering timer for 0x7f759cb26218 (ticks=908866499, timeout=908866499 (0 s), wr_timeout=0 (211631299 s)), write queue: 0 bytes In tcpdump I can see that the connection does not get closed immediately. Is there some timer in the tcp stack which cleans up expired connections from time to time? Can I manipulate how often this timer runs? I couldn't find anything, and clearly even though I tell Kamailio to close the connection after 1 second, it doesn't do it. Thanks for any input. Regards, Sebastian