Reloading config no longer crash container but do not reload config

[jrgn9]

Hi again!

Ref #21, I just updated to the latest build. The reload no longer crashes/kills the container, but it doesn't seem like the config gets reloaded either.

I spun up the container, edited the log format in haproxy.cfg (so it would be easy to see that a config reload has happened), then ran docker kill -s USR2 haproxy

The log says that it reloads, and I see that all the backends/frontends stops, then start up again. This is the log after running USR2 kill:

[Logs with current format]

[NOTICE]   (20) : Reloading HAProxy
time="2026-09-28T07:26:27Z" level=warning msg="event listener disconnected, reconnecting: could not listen to HAProxy's events: EventListener(dpapi): dial unix /var/run/haproxy-master.sock: connect: connection refused"
[NOTICE]   (20) : Initializing new worker (48)
[NOTICE]   (20) : haproxy version is 3.4.5-c28c410
[WARNING]  (20) : Former worker (22) exited with code 0 (Exit)
Proxy frontend1 stopped (cumulated conns: FE: 9, BE: 0).
[NOTICE]   (20) : Loading success.
[WARNING]  (43) : Proxy frontend1 stopped (cumulated conns: FE: 9, BE: 0).
[WARNING]  (43) : Proxy frontend2 stopped (cumulated conns: FE: 0, BE: 0).
[WARNING]  (43) : Proxy frontend3 stopped (cumulated conns: FE: 0, BE: 0).
[WARNING]  (43) : Proxy frontend4 stopped (cumulated conns: FE: 0, BE: 0).
[WARNING]  (43) : Proxy frontend5 stopped (cumulated conns: FE: 0, BE: 0).
[WARNING]  (43) : Proxy frontend6 stopped (cumulated conns: FE: 1, BE: 0).
[WARNING]  (43) : Proxy frontend7 stopped (cumulated conns: FE: 6, BE: 0).
[WARNING]  (43) : Proxy backend1 stopped (cumulated conns: FE: 0, BE: 0).
Proxy frontend8 stopped (cumulated conns: FE: 0, BE: 0).
[WARNING]  (43) : Proxy backend2 stopped (cumulated conns: FE: 0, BE: 0).
[WARNING]  (43) : Proxy backend3 stopped (cumulated conns: FE: 0, BE: 1).
Proxy frontend9 stopped (cumulated conns: FE: 0, BE: 0).
Proxy frontend5 stopped (cumulated conns: FE: 0, BE: 0).
Proxy frontend6 stopped (cumulated conns: FE: 0, BE: 0).
Proxy frontend7 stopped (cumulated conns: FE: 1, BE: 0).
Proxy backend2 stopped (cumulated conns: FE: 6, BE: 0).
Proxy backend3 stopped (cumulated conns: FE: 0, BE: 0).
Proxy backend4 stopped (cumulated conns: FE: 0, BE: 0).
Proxy backend2 stopped (cumulated conns: FE: 0, BE: 1).
Proxy backend5 stopped (cumulated conns: FE: 0, BE: 0).
Proxy backend4 stopped (cumulated conns: FE: 0, BE: 6).
Proxy backend6 stopped (cumulated conns: FE: 0, BE: 3).
Proxy backend7 stopped (cumulated conns: FE: 0, BE: 2).
Proxy backend8 stopped (cumulated conns: FE: 0, BE: 2).
[WARNING]  (43) : Proxy backend3 stopped (cumulated conns: FE: 0, BE: 0).
[WARNING]  (43) : Proxy backend9 stopped (cumulated conns: FE: 0, BE: 6).
[WARNING]  (43) : Proxy backend9 stopped (cumulated conns: FE: 0, BE: 3).
[WARNING]  (43) : Proxy backend10 stopped (cumulated conns: FE: 0, BE: 2).
[WARNING]  (43) : Proxy backend9 stopped (cumulated conns: FE: 0, BE: 2).
[WARNING]  (43) : Proxy backend11 stopped (cumulated conns: FE: 0, BE: 2).
Proxy backend11 stopped (cumulated conns: FE: 0, BE: 2).

[proceeds as usual with same log format]

Then running docker restart haproxy restarts the container and the log format is changed (config reload).

I now know that swapping out the s6 layer is on it's way, so I understand if a fix for this is not a priority/will not be fixed. I just thought an update on the issue was in place.

[dkorunic]

@jrgn9 Reloading through SIGUSR2 (gopherd doing signal-forwarding) works given you have observed Reloading HAProxy log line. It will take effect for new connections, while the old connections (one that have been established prior to configuration reload) will still be processed with the old settings.
The other important point here is that HAProxy configuration in s6/gopherd enabled-images is usually managed through Dataplaneapi and should not be rewritten by hand, as long as Dataplaneapi is active.