move embedded haproxy process to a distinct pid group

The embedded haproxy process is being created in the same pid group of
the controller, which makes it receive the same signals received by the
controller when running under a process supervisor. This makes haproxy
be prematurely terminated when the controller is still shutting down.

builder/Dockerfile

@@ -44,4 +44,7 @@ RUN mkdir -p /var/empty /etc/haproxy /var/lib/haproxy /var/run/haproxy\
 
 STOPSIGNAL SIGTERM
 USER haproxy
+
+# dumb-init reaps the old haproxy process in the embedded, non master-worker mode,
+# after receiving SIGUSR1, avoiding it to become a zombie.
 ENTRYPOINT ["/usr/bin/dumb-init", "--", "/start.sh"]

pkg/haproxy/instance.go

@@ -637,6 +637,7 @@ func (i *instance) startHAProxySync() {
 		"-f", i.options.HAProxyCfgDir)
 	cmd.Stdout = os.Stdout
 	cmd.Stderr = os.Stderr
+	cmd.SysProcAttr = &syscall.SysProcAttr{Setpgid: true}
 	if err := cmd.Start(); err != nil {
 		i.logger.Error("error starting haproxy: %v", err)
 		return

rootfs/Dockerfile

@@ -27,4 +27,7 @@ RUN mkdir -p /var/empty /etc/haproxy /var/lib/haproxy /var/run/haproxy\
 
 STOPSIGNAL SIGTERM
 USER haproxy
+
+# dumb-init reaps the old haproxy process in the embedded, non master-worker mode,
+# after receiving SIGUSR1, avoiding it to become a zombie.
 ENTRYPOINT ["/usr/bin/dumb-init", "--", "/start.sh"]