Add healthz endpoints and container healthchecks

backend/Dockerfile

@@ -69,6 +69,9 @@ RUN python3 manage.py collectstatic --noinput --verbosity 2
 # Expose ports
 EXPOSE 80 8000
 
+HEALTHCHECK --interval=30s --timeout=5s --start-period=60s --retries=5 \
+  CMD ["python3", "/code/healthcheck.py"]
+
 # Start with an entrypoint that runs init tasks then starts supervisord
 ENTRYPOINT ["/code/entrypoint.sh"]
 

backend/server/healthcheck.py

@@ -0,0 +1,30 @@
+#!/usr/bin/env python3
+import os
+import sys
+import urllib.error
+import urllib.request
+
+
+def main() -> int:
+    timeout = float(os.getenv('HEALTHCHECK_TIMEOUT_SECONDS', '3'))
+    urls = os.getenv(
+        'HEALTHCHECK_URLS',
+        'http://127.0.0.1:8000/healthz,http://127.0.0.1/healthz',
+    )
+
+    for raw_url in urls.split(','):
+        url = raw_url.strip()
+        if not url:
+            continue
+        try:
+            with urllib.request.urlopen(url, timeout=timeout) as response:
+                if response.status == 200:
+                    return 0
+        except (OSError, urllib.error.URLError):
+            continue
+
+    return 1
+
+
+if __name__ == '__main__':
+    sys.exit(main())

backend/server/main/urls.py

@@ -2,7 +2,7 @@
 from django.contrib import admin
 from django.views.generic import RedirectView, TemplateView
 from users.views import IsRegistrationDisabled, PublicUserListView, PublicUserDetailView, UserMetadataView, UpdateUserMetadataView, EnabledSocialProvidersView, DisablePasswordAuthenticationView
-from .views import get_csrf_token, get_public_url, serve_protected_media
+from .views import get_csrf_token, get_public_url, healthz, serve_protected_media
 from drf_yasg.views import get_schema_view
 from drf_yasg import openapi
 
@@ -13,6 +13,7 @@
     )
 )
 urlpatterns = [
+    path('healthz/', healthz, name='healthz'),
     path('api/', include('adventures.urls')),
     path('api/', include('worldtravel.urls')),
     path("auth/", include("allauth.headless.urls")),
@@ -49,4 +50,4 @@
     path("api/integrations/", include("integrations.urls")),
 
     # Include the API endpoints:   
-]
+]

backend/server/main/views.py

@@ -6,6 +6,11 @@
 from django.views.static import serve
 from adventures.utils.file_permissions import checkFilePermission
 
+
+def healthz(request):
+    return HttpResponse('ok', content_type='text/plain')
+
+
 def get_csrf_token(request):
     csrf_token = get_token(request)
     return JsonResponse({'csrfToken': csrf_token})
@@ -39,4 +44,4 @@ def serve_protected_media(request, path):
             response = HttpResponse()
             response['Content-Type'] = ''
             response['X-Accel-Redirect'] = '/protectedMedia/' + path
-            return response
+            return response

docker-compose-traefik.yaml

@@ -43,7 +43,14 @@ services:
       - "traefik.http.routers.adventurelogweb.tls=true"
       - "traefik.http.routers.adventurelogweb.tls.certresolver=letsencrypt"
     depends_on:
-      - server
+      server:
+        condition: service_healthy
+    healthcheck:
+      test: ["CMD", "node", "/app/healthcheck.mjs"]
+      interval: 30s
+      timeout: 5s
+      retries: 5
+      start_period: 30s
 
   server:
     image: ghcr.io/seanmorley15/adventurelog-backend:latest
@@ -71,6 +78,12 @@ services:
       - "traefik.http.routers.adventurelogserver.tls.certresolver=letsencrypt"
     depends_on:
       - db
+    healthcheck:
+      test: ["CMD", "python3", "/code/healthcheck.py"]
+      interval: 30s
+      timeout: 5s
+      retries: 5
+      start_period: 60s
 
 volumes:
   postgres-data:

docker-compose.dev.yml

@@ -12,11 +12,18 @@ services:
     ports:
       - "${FRONTEND_PORT:-8015}:3000"
     depends_on:
-      - server
+      server:
+        condition: service_healthy
     volumes:
       - ./frontend:/app
       - pnpm_store:/pnpm-store
     command: sh -c "mkdir -p /pnpm-store && chown -R node:node /pnpm-store && su node -c 'pnpm config set store-dir /pnpm-store && pnpm install --frozen-lockfile && pnpm exec vite dev --host 0.0.0.0 --port 3000 --strictPort'"
+    healthcheck:
+      test: ["CMD", "node", "/app/healthcheck.mjs"]
+      interval: 30s
+      timeout: 5s
+      retries: 5
+      start_period: 30s
 
   db:
     image: postgis/postgis:16-3.5
@@ -52,6 +59,12 @@ services:
         python manage.py createsuperuser --noinput --username \"$$DJANGO_SUPERUSER_USERNAME\" --email \"$$DJANGO_SUPERUSER_EMAIL\" || true;
       fi;
       python manage.py runserver 0.0.0.0:8000"
+    healthcheck:
+      test: ["CMD", "python3", "/code/healthcheck.py"]
+      interval: 30s
+      timeout: 5s
+      retries: 5
+      start_period: 60s
 
 volumes:
   postgres_data:

docker-compose.yml

@@ -8,7 +8,14 @@ services:
     ports:
       - "${FRONTEND_PORT:-8015}:3000"
     depends_on:
-      - server
+      server:
+        condition: service_healthy
+    healthcheck:
+      test: ["CMD", "node", "/app/healthcheck.mjs"]
+      interval: 30s
+      timeout: 5s
+      retries: 5
+      start_period: 30s
 
   db:
     image: postgis/postgis:16-3.5
@@ -30,6 +37,12 @@ services:
       - db
     volumes:
       - adventurelog_media:/code/media/
+    healthcheck:
+      test: ["CMD", "python3", "/code/healthcheck.py"]
+      interval: 30s
+      timeout: 5s
+      retries: 5
+      start_period: 60s
 
 volumes:
   postgres_data:

frontend/Dockerfile

@@ -45,5 +45,8 @@ USER node:node
 # Expose the port that the app is listening on
 EXPOSE 3000
 
+HEALTHCHECK --interval=30s --timeout=5s --start-period=30s --retries=5 \
+  CMD ["node", "/app/healthcheck.mjs"]
+
 # Run startup.sh instead of the default command
-CMD ["./startup.sh"]
+CMD ["./startup.sh"]

frontend/healthcheck.mjs

@@ -0,0 +1,27 @@
+const timeoutMs = Number.parseInt(process.env.HEALTHCHECK_TIMEOUT_MS ?? '3000', 10);
+const urls = (process.env.HEALTHCHECK_URLS ?? 'http://127.0.0.1:3000/healthz')
+	.split(',')
+	.map((url) => url.trim())
+	.filter(Boolean);
+
+for (const url of urls) {
+	const controller = new AbortController();
+	const timer = setTimeout(() => controller.abort(), timeoutMs);
+
+	try {
+		const response = await fetch(url, {
+			method: 'GET',
+			signal: controller.signal,
+			cache: 'no-store'
+		});
+		if (response.status === 200) {
+			process.exit(0);
+		}
+	} catch (_) {
+		// Continue to the next URL.
+	} finally {
+		clearTimeout(timer);
+	}
+}
+
+process.exit(1);

frontend/src/routes/healthz/+server.ts

@@ -0,0 +1,10 @@
+import type { RequestHandler } from './$types';
+
+export const GET: RequestHandler = async () => {
+	return new Response('ok', {
+		status: 200,
+		headers: {
+			'Content-Type': 'text/plain'
+		}
+	});
+};