opencloud_full: add healthchecks for traefik, opencloud and stalwart to the docker compose project

devtools/deployments/opencloud_full/docker-compose.yml

@@ -37,6 +37,8 @@ services:
       - "--accessLog.format=json"
       - "--accessLog.fields.headers.names.X-Request-Id=keep"
       - "--accessLog.fields.headers.names.Trace-Id=keep"
+      # enable the ping endpoint for monitoring
+      - "--ping=true"
     ports:
       - "80:80"
       - "443:443"
@@ -55,6 +57,12 @@ services:
     logging:
       driver: ${LOG_DRIVER:-local}
     restart: always
+    healthcheck:
+      test: ["CMD", "traefik", "healthcheck", "--ping=true"]
+      interval: 10s
+      retries: 5
+      start_period: 3s
+      timeout: 3s
 
 volumes:
   certs:

devtools/deployments/opencloud_full/opencloud.yml

@@ -21,7 +21,7 @@ services:
       # enable services that are not started automatically
       OC_ADD_RUN_SERVICES: ${START_ADDITIONAL_SERVICES}
       OC_URL: https://${OC_DOMAIN:-cloud.opencloud.test}
-      OC_LOG_LEVEL: ${LOG_LEVEL:-info}
+      OC_LOG_LEVEL: "${LOG_LEVEL:-info}"
       OC_LOG_COLOR: "${LOG_PRETTY:-false}"
       OC_LOG_PRETTY: "${LOG_PRETTY:-false}"
       # do not use SSL between Traefik and OpenCloud
@@ -88,6 +88,12 @@ services:
     logging:
       driver: ${LOG_DRIVER:-local}
     restart: always
+    healthcheck:
+      test: ["CMD-SHELL", "test $(curl -sSf http://localhost:9104/healthz) == 'OK'"]
+      interval: 10s
+      retries: 5
+      start_period: 3s
+      timeout: 3s
 
   opencloud-certs:
     image: alpine/openssl:latest

devtools/deployments/opencloud_full/stalwart.yml

@@ -48,6 +48,13 @@ services:
         echo "Starting Stalwart in production mode"
         exec /usr/local/bin/stalwart --config /etc/stalwart/config.json
 
+    healthcheck:
+      test: ["CMD-SHELL", "curl -LsSf http://localhost:8080/healthz/live|grep -q '\"status\":200'"]
+      interval: 10s
+      retries: 5
+      start_period: 3s
+      timeout: 3s
+
   stalwart-import:
     image: stalwart-cli:latest
     build: