Add option to autostart managed containers (#43)

This is useful for Podman, where containers are not autostarted on
boot like they are under the Docker daemon. The usual way to solve this
problem under Podman is to use systemd, but it's annoying to create a
systemd unit for every ExApps-managed container by hand.

Signed-off-by: AJ Jordan <alex@strugee.net>

Author: strugee

Dockerfile

@@ -23,7 +23,8 @@ RUN set -ex; \
         bind-tools \
         nano \
         vim \
-        envsubst; \
+        envsubst \
+        jq; \
     chmod -R 777 /tmp
 
 COPY --chmod=775 *.sh /

README.md

@@ -75,6 +75,8 @@ You should set `BIND_ADDRESS` to the IP on which server with ExApps can accept r
 
 `TIMEOUT_SERVER`: timeout for ExApp to start responding to NC request, default: **1800s**
 
+`NC_AUTOSTART_CONTAINERS`: if set, automatically start managed app containers (useful for Podman users)
+
 `NC_HAPROXY_PASSWORD_FILE`: Specifies path to a file containing the password for HAProxy. 
 
 > [!NOTE]

start.sh

@@ -52,12 +52,21 @@ fi
 
 echo "HaProxy config:"
 
+start_app_containers() {
+  if ! [ -z ${NC_AUTOSTART_CONTAINERS+x} ]; then
+    echo "Autostarting existing app containers per NC_AUTOSTART_CONTAINERS."
+    curl --silent --globoff -XGET --unix-socket /run/docker.sock 'http://localhost/containers/json?filters={"status":["exited"]}' | jq -r '.[] | select(.Names[0] | startswith("/nc_app_")) | .Id' | xargs -I% curl --silent --globoff -XPOST --fail-with-body --unix-socket /run/docker.sock 'http://localhost/containers/%/start'
+  fi
+}
+
 if [ -f "/certs/cert.pem" ]; then
   cat /haproxy.cfg
   cat /haproxy_ex_apps.cfg
+  start_app_containers
   haproxy -f /haproxy.cfg -f /haproxy_ex_apps.cfg -db
 else
   cat /haproxy.cfg
+  start_app_containers
   haproxy -f /haproxy.cfg -db
 fi