Hello,
As for configuration, yes, as Lukas pointed, "hard-stop-after" is a
right setting, it was designed specifically for such case of long
connections.
If master-worker mode is used, there is also "mworker-max-reloads",
which can help.
https://docs.haproxy.org/3.0/configuration.html#3.1-mworker-max-reloads
Basically, you need to set "mworker-max-reloads" as some small value N
(2-3 times). If the old worker hasn't quit up to N reloads, master
process will send a SIGTERM to it during the next reload (N+1). It could
be handy to control the old worker processes number.
It can be used in combination with "hard-stop-after".
--
Bien cordialement,
Valentine
