Fix nginx issues #2

.forgejo/workflows/ci.yml

@@ -230,60 +230,52 @@ jobs:
       PROD_BACKEND_HOST: ${{ secrets.PROD_BACKEND_HOST }}
 
     steps:
-      - name: Checkout code
-        uses: actions/checkout@v4
+    - name: Checkout code
+      uses: actions/checkout@v4
 
-      - name: Login to container registry with PAT
-        run: |
-          echo "${{ secrets.REGISTRY_TOKEN }}" | podman --remote login \
-            -u "${{ secrets.REGISTRY_USERNAME }}" \
-            --password-stdin \
-            "${{ secrets.REGISTRY_HOST }}"
+    - name: Login to container registry with PAT
+      run: |
+        echo "${{ secrets.REGISTRY_TOKEN }}" | podman --remote login \
+          -u "${{ secrets.REGISTRY_USERNAME }}" \
+          --password-stdin \
+          "${{ secrets.REGISTRY_HOST }}"
 
-      - name: Pull production images (optional but faster on play)
-        run: |
-          podman --remote pull "$REGISTRY_HOST/$APP_NAME/sharenet-backend-api-postgres:$IMAGE_TAG"
-          podman --remote pull "$REGISTRY_HOST/$APP_NAME/sharenet-frontend:$IMAGE_TAG"
+    - name: (Optional) Pre-pull images to speed up play
+      run: |
+        podman --remote pull "$REGISTRY_HOST/$APP_NAME/sharenet-backend-api-postgres:$IMAGE_TAG"
+        podman --remote pull "$REGISTRY_HOST/$APP_NAME/sharenet-frontend:$IMAGE_TAG"
 
-      # RENDER nginx.conf FROM REPO AND COPY TO HOST (no unshare)
-      - name: Render nginx.conf and write to host
-        run: |
-          set -euo pipefail
-          apk add --no-cache gettext >/dev/null  # provides envsubst
-          # Render with your CI env (PROD_* vars)
-          envsubst < nginx/nginx.conf > /tmp/nginx.conf
+    # 1) Render nginx.conf from repo and copy to host (no unshare)
+    - name: Render nginx.conf and write to host
+      run: |
+        set -euo pipefail
+        apk add --no-cache gettext >/dev/null
+        envsubst < nginx/nginx.conf > /tmp/nginx.conf
+        podman --remote run --rm -i \
+          --userns=keep-id \
+          -v /opt/sharenet/nginx:/host-nginx:rw \
+          alpine:3.20 sh -c 'install -D -m 0644 /dev/stdin /host-nginx/nginx.conf' \
+          < /tmp/nginx.conf
 
-          # Copy to host via remote Podman bind-mount; keep prod-service uid/gid
-          podman --remote run --rm -i \
-            --userns=keep-id \
-            -v /opt/sharenet/nginx:/host-nginx:rw \
-            alpine:3.20 sh -c 'install -D -m 0644 /dev/stdin /host-nginx/nginx.conf' \
-            < /tmp/nginx.conf
+    # 2) Lint nginx.conf BEFORE restarting the pod (avoids crash loops)
+    - name: Validate nginx.conf with throwaway container
+      run: |
+        set -euo pipefail
+        podman --remote run --rm \
+          -v /opt/sharenet/nginx:/etc/nginx:ro \
+          docker.io/nginx:alpine \
+          sh -lc 'nginx -t -c /etc/nginx/nginx.conf'
 
-      # TRY ZERO-DOWNTIME RELOAD FIRST
-      - name: Reload in-pod Nginx (or restart on failure)
-        continue-on-error: true
-        run: |
-          set -euo pipefail
-          podman --remote exec sharenet-production-pod-nginx nginx -t
-          podman --remote exec sharenet-production-pod-nginx nginx -s reload
+    # 3) Recreate the pod (down → play). This ensures new images/ports/env + the validated conf.
+    - name: Recreate pod
+      run: |
+        set -euo pipefail
+        podman --remote kube down sharenet-production-pod || true
+        envsubst < deploy/prod-pod.yml | podman --remote kube play -
 
-      - name: Fallback restart Nginx container if reload failed
-        if: failure()
-        run: |
-          set -euo pipefail
-          podman --remote restart sharenet-production-pod-nginx
+    # 4) Verify health on the prod host
+    - name: Verify in-pod Nginx
+      run: |
+        set -euo pipefail
+        curl -sS -D- http://127.0.0.1:18080/healthz
 
-      # (Re)APPLY THE POD (ensures new images/config picked up)
-      - name: Recreate pod (down & play)
-        run: |
-          set -euo pipefail
-          podman --remote kube down sharenet-production-pod || true
-          # Render your pod manifest (uses same $ENV as before)
-          envsubst < deploy/prod-pod.yml | podman --remote kube play -
-
-      # VERIFY
-      - name: Verify in-pod nginx is healthy
-        run: |
-          set -euo pipefail
-          curl -sS -D- http://127.0.0.1:18080/healthz