Hi, I don't know if it could help, but based on Thomas instructions/example, I'm able to reproduce:
As weird as it may seem, the 'maxconn' parameter used with dynamic server seems to trigger the issue. Below is a minimal reproducer: > global > stats socket /tmp/ha.sock mode 660 level admin expose-fd listeners > > defaults > timeout client 5s > timeout server 5s > timeout connect 5s > > frontend test > mode http > bind *:8081 > use_backend farm > > listen dummyok > bind localhost:18999 > mode http > http-request return status 200 hdr test "ok" > > backend farm > mode http (it doesn't matter if the backend starts with empty server list) Step 1 (optional), test the reference behavior without maxconn: Start haproxy and perform the following : echo "add server farm/t1 127.0.0.1:18999" | nc -U /tmp/ha.sock echo "enable server farm/t1" | nc -U /tmp/ha.sock curl localhost:8081 -> OK Step 2, compare behavior when maxconn is in play: Start haproxy and perform the following : echo "add server farm/t1 127.0.0.1:18999 maxconn 100" | nc -U /tmp/ha.sock echo "enable server farm/t1" | nc -U /tmp/ha.sock curl localhost:8081 -> 503 after 5s connect timeout Regards, Aurelien