Blue green

.github/workflows/ci.yml

@@ -124,37 +124,40 @@ jobs:
           echo 'CONN_MAX_AGE=60' >> docker.prod.env
           echo 'CONN_HEALTH_CHECKS="true"' >> docker.prod.env
           echo 'SANDBOX_DISABLE_PROC="true"' >> docker.prod.env
-          echo 'ALLOWED_HOSTS="backend,localhost,127.0.0.1"' >> docker.prod.env
+          echo 'ALLOWED_HOSTS="backend,backend-green,backend-blue,localhost,127.0.0.1"' >> docker.prod.env
           echo 'USE_SANDBOX_JAIL="on"' >> docker.prod.env
           echo 'CI=true' >> docker.prod.env
 
       - name: Comment out SSL server configuration from nginx
         run: |
-          sed -i '/{{HTTPS_SERVER_BLOCK_START}}/,/{{HTTPS_SERVER_BLOCK_END}}/s/^/#/' nginx/production.conf
+          sed -i '/{{HTTPS_SERVER_BLOCK_START}}/,/{{HTTPS_SERVER_BLOCK_END}}/s/^/#/' nginx/production/default.conf
 
       - name: Bring up postgres and nginx containers
         run: |
+          cp ./nginx/production/runtime/geo.conf.example      ./nginx/production/runtime/geo.conf
+          cp ./nginx/production/runtime/upstream.conf.example ./nginx/production/runtime/upstream.conf
+          docker compose -f docker-compose.prod.yaml build nginx
           docker compose -f docker-compose.prod.yaml up -d postgres nginx
           timeout 15s docker compose -f docker-compose.prod.yaml logs -f || true
           docker compose logs nginx | grep "ready for start up"
           ! docker compose logs nginx | grep -q "nginx-1 exited with code"
 
       - name: Build and bring up up backend container
         run: |
-          docker compose -f docker-compose.prod.yaml build backend
-          docker compose -f docker-compose.prod.yaml up -d backend
+          docker compose -f docker-compose.prod.yaml build backend-blue
+          docker compose -f docker-compose.prod.yaml up -d backend-blue
           timeout 15s docker compose -f docker-compose.prod.yaml logs -f || true
 
       - name: Build and bring up up frontend container
         run: |
-          docker compose -f docker-compose.prod.yaml build frontend
-          docker compose -f docker-compose.prod.yaml up -d frontend
+          docker compose -f docker-compose.prod.yaml build frontend-blue
+          docker compose -f docker-compose.prod.yaml up -d frontend-blue
           timeout 15s docker compose -f docker-compose.prod.yaml logs -f || true
 
       - name: Sanity check the endpoints
         run: |
-          curl --silent http://localhost:8080/ | head -c 256
-          curl --silent http://localhost:8000/api/ | jq
+          curl --silent http://localhost/healthz | jq
+          curl --silent http://localhost/api/healthz | jq
 
       - name: Shut everything down
         run: |

.gitignore

@@ -25,3 +25,6 @@ docker.prod.env
 send_update.py
 update.sh
 .venv/
+nginx/production/runtime/*.conf
+!nginx/production/runtime/*.conf.example
+.deploy.env

backend/coreapp/middleware.py

@@ -45,6 +45,7 @@ def middleware(request: HttpRequest) -> Response:
 def is_public_get_request(req: Request) -> bool:
     public_paths = [
         "/api/compiler",
+        "/api/healthz$",
         "/api/library",
         "/api/platform",
         "/api/preset",

backend/coreapp/tests/test_request.py

@@ -12,6 +12,14 @@
 
 
 class RequestTests(APITestCase):
+    def test_health_check_is_stateless(self) -> None:
+        response = self.client.get(reverse("healthz"), HTTP_USER_AGENT="browser")
+
+        self.assertEqual(response.status_code, status.HTTP_200_OK)
+        self.assertEqual(response.json(), {"ok": True})
+        self.assertEqual(Profile.objects.count(), 0)
+        self.assertNotIn("sessionid", response.cookies)
+
     def test_cookie_less_current_user_does_not_create_profile(self) -> None:
         """
         Ensure that a passive current-user read does not create a session profile.

backend/coreapp/urls.py

@@ -3,6 +3,7 @@
 
 from coreapp.views import (
     compiler,
+    health,
     library,
     platform,
     preset,
@@ -22,6 +23,7 @@
 urlpatterns = [
     *router.urls,
     path("compiler", compiler.CompilerDetail.as_view(), name="compiler"),
+    path("healthz", health.HealthCheck.as_view(), name="healthz"),
     path(
         "compiler/<str:platform>/<str:compiler>",
         compiler.SingleCompilerDetail.as_view(),

backend/coreapp/views/health.py

@@ -0,0 +1,15 @@
+from django.db import connection
+from rest_framework import status
+from rest_framework.request import Request
+from rest_framework.response import Response
+from rest_framework.views import APIView
+
+
+class HealthCheck(APIView):
+    def get(self, request: Request) -> Response:
+        try:
+            connection.ensure_connection()
+        except Exception:
+            return Response({"ok": False}, status=status.HTTP_503_SERVICE_UNAVAILABLE)
+
+        return Response({"ok": True})

backend/docker_prod_entrypoint.sh

@@ -13,8 +13,6 @@ until nc -z ${DB_HOST} ${DB_PORT} > /dev/null; do
   sleep 1
 done
 
-uv run /backend/manage.py migrate
-
 if [ -z "$CI" ]; then
   uv run manage.py clearsessions
   uv run /backend/housekeeping.py