From cf18565b80995dcb35499b9c6b8450dfde120312 Mon Sep 17 00:00:00 2001
From: Alberto Bertogli <albertito@blitiri.com.ar>
Date: Mon, 4 Aug 2025 11:11:50 +0100
Subject: [PATCH] docker: Use supervisord to launch chasquid and dovecot

Today, we launch dovecot in the background and chasquid in the
foreground using sudo.

This means that dovecot failures won't propagate, and signals to the
container (e.g. to stop it) also don't get propagated to dovecot
(because it's in the background) or chasquid (because they don't go
beyond the sudo process).

Thanks to [Guiorgy@github](https://github.com/Guiorgy) for identifying
the problem, proposing alternatives, help debugging, and discussing this
in https://github.com/albertito/chasquid/pull/70.
---
 docker/Dockerfile       |  4 ++-
 docker/entrypoint.sh    | 10 +++----
 docker/supervisord.conf | 63 +++++++++++++++++++++++++++++++++++++++++
 3 files changed, 70 insertions(+), 7 deletions(-)
 create mode 100644 docker/supervisord.conf

diff --git a/docker/Dockerfile b/docker/Dockerfile
index e5d97d9..ef1ebd0 100644
--- a/docker/Dockerfile
+++ b/docker/Dockerfile
@@ -27,7 +27,8 @@ RUN DEBIAN_FRONTEND=noninteractive \
 		chasquid \
 		dovecot-lmtpd dovecot-imapd dovecot-pop3d \
 		dovecot-sieve dovecot-managesieved \
-		acl libcap2-bin sudo certbot && \
+		supervisor \
+		acl libcap2-bin certbot && \
 	apt-get autoremove --purge -y -q && \
 	apt-get autoclean -y -q && \
 	rm -rf /var/lib/apt/lists/*
@@ -42,6 +43,7 @@ RUN setcap CAP_NET_BIND_SERVICE=+eip /usr/bin/chasquid
 # Copy docker-specific configurations.
 COPY docker/dovecot.conf /etc/dovecot/dovecot.conf
 COPY docker/chasquid.conf /etc/chasquid/chasquid.conf
+COPY docker/supervisord.conf /etc/supervisor/supervisord.conf
 
 # Copy utility scripts.
 COPY docker/add-user.sh docker/entrypoint.sh /
diff --git a/docker/entrypoint.sh b/docker/entrypoint.sh
index 5302169..dd6873f 100755
--- a/docker/entrypoint.sh
+++ b/docker/entrypoint.sh
@@ -102,9 +102,7 @@ done >> /etc/dovecot/auto-ssl.conf
 sed -i '/^hostname:/d' /etc/chasquid/chasquid.conf
 echo "hostname: '$ONE_DOMAIN'" >> /etc/chasquid/chasquid.conf
 
-# Start the services: dovecot in background, chasquid in foreground.
-start-stop-daemon --start --quiet --pidfile /run/dovecot.pid \
-	--exec /usr/sbin/dovecot -- -c /etc/dovecot/dovecot.conf
-
-# shellcheck disable=SC2086
-sudo -u chasquid -g chasquid  /usr/bin/chasquid  $CHASQUID_FLAGS
+# Start the services (dovecot and chasquid, configured in supervisord.conf).
+# We exec, so supervisord becomes our init, and it forwards any signals we
+# receive from outside the container.
+exec supervisord --nodaemon -c /etc/supervisor/supervisord.conf
diff --git a/docker/supervisord.conf b/docker/supervisord.conf
new file mode 100644
index 0000000..e9bf42a
--- /dev/null
+++ b/docker/supervisord.conf
@@ -0,0 +1,63 @@
+
+# We launch two daemons inside the container: chasquid and dovecot.
+# The supervisord program will act as init, and forward signals to them from
+# outside the container, so shutdown/restarts/etc. is handled appropriately.
+
+[supervisord]
+user = root
+# We run with --nodaemon (as per entrypoint.sh), that implicitly makes
+# supervisord log to stdout _in addition_ to logfile.
+# So set logfile to /dev/null to avoid duplicated entries.
+logfile = /dev/null
+logfile_maxbytes = 0
+
+[program:dovecot]
+command = /usr/sbin/dovecot -F -c /etc/dovecot/dovecot.conf
+# TODO: confirm we don't need stopasgroup, or add it.
+stdout_logfile = /dev/stdout
+stdout_logfile_maxbytes = 0
+stderr_logfile = /dev/stderr
+stderr_logfile_maxbytes = 0
+autorestart = false
+priority = 201
+
+[program:chasquid]
+command = /usr/bin/chasquid
+user = chasquid
+group = chasquid
+environment = USER="chasquid"
+stdout_logfile = /dev/stdout
+stdout_logfile_maxbytes = 0
+stderr_logfile = /dev/stderr
+stderr_logfile_maxbytes = 0
+autorestart = false
+# chasquid doesn't _need_ dovecot to start earlier, but it will emit a warning
+# if dovecot auth it's not ready at startup. To minimize the chances of
+# problems, have supervisord start it after dovecot. Still it's not guaranteed
+# to be fully up by the time chasquid needs it (because it starts the
+# components asynchronously), but it minimizes the problem.
+priority = 202
+
+
+# We intentionally don't auto-restart subprocesses on unexpected crashes, and
+# make supervisord exit if that happens. That way, whatever error caused it is
+# propagated to the container runner and the user can monitor and decide what
+# to do.
+[eventlistener:exit_on_process_fatal]
+# fatal: when the starting attempts fail (e.g. bad configuration).
+# exited: when it fails after starting up (e.g. random crash).
+# stopped: when a user wants to stop the container (e.g. requests a restart,
+#   or a manual stop). We don't interfere with this one because it's
+#   intentionally targeting a specific process. If the user wanted to stop the
+#   whole container, they can.
+events = PROCESS_STATE_FATAL, PROCESS_STATE_EXITED
+priority = 100
+command = /bin/bash -c '
+	while true; do
+		echo "READY";
+		read line;
+		kill -SIGTERM $PPID;
+		echo "RESULT 2";
+		echo -n "OK";
+	done;
+	'