Deploy to production: GitHub Actions + ghcr.io + Kubernetes

- Switch from Gitea to GitHub Container Registry (ghcr.io) - Add GitHub Actions workflow with Tailscale connectivity - Update k8s manifests for cloud nodes and Traefik ingress - Configure for turbo.kazcloud.dev domain - Test deployment with home page text change
2026-04-16 22:12:53 +00:00 · 2026-03-29 08:46:27 -04:00
parent 2bb1dfa1e4
commit 69993a3bf5
14 changed files with 793 additions and 596 deletions
--- a/TODO.md
+++ b/TODO.md
@@ -0,0 +1,130 @@
+# TurboVault TODO
+
+## Fix Gitea Ingress for Large Container Pushes
+
+**Issue:** Docker push to Gitea registry times out on large layers (~250MB)
+
+**Error:** 
+```
+Package registry API internal error: 500 unexpected EOF
+```
+
+**Root Cause:** Ingress controller (nginx/traefik) in front of Gitea has timeout limits that prevent large uploads from completing.
+
+**Symptoms:**
+- Small layers push fine (already exists)
+- Large gem layer (12f753f9ec10, ~250MB) times out after ~1 minute
+- Gitea logs show: `PATCH /v2/ryan/turbovault-app/blobs/uploads/... elapsed 3275.4ms ... 500 unexpected EOF`
+
+### Solution: Update Gitea Ingress
+
+Find your Gitea ingress configuration and add these annotations:
+
+**For NGINX Ingress Controller:**
+```yaml
+apiVersion: networking.k8s.io/v1
+kind: Ingress
+metadata:
+  name: gitea
+  namespace: tools  # or wherever gitea is
+  annotations:
+    # Allow unlimited upload size
+    nginx.ingress.kubernetes.io/proxy-body-size: "0"
+    
+    # Increase timeouts to 10 minutes
+    nginx.ingress.kubernetes.io/proxy-read-timeout: "600"
+    nginx.ingress.kubernetes.io/proxy-send-timeout: "600"
+    nginx.ingress.kubernetes.io/proxy-connect-timeout: "600"
+    
+    # Enable chunked uploads
+    nginx.ingress.kubernetes.io/proxy-request-buffering: "off"
+```
+
+**For Traefik Ingress:**
+```yaml
+apiVersion: traefik.containo.us/v1alpha1
+kind: Middleware
+metadata:
+  name: gitea-buffering
+  namespace: tools
+spec:
+  buffering:
+    maxRequestBodyBytes: 0  # Unlimited
+    memRequestBodyBytes: 2097152  # 2MB in memory, rest to disk
+
+---
+apiVersion: networking.k8s.io/v1
+kind: Ingress
+metadata:
+  name: gitea
+  namespace: tools
+  annotations:
+    traefik.ingress.kubernetes.io/router.middlewares: tools-gitea-buffering@kubernetescrd
+```
+
+### Steps to Fix:
+
+1. **Find current Gitea ingress:**
+   ```bash
+   kubectl get ingress -n tools
+   kubectl get ingress gitea -n tools -o yaml
+   ```
+
+2. **Identify ingress controller:**
+   ```bash
+   kubectl get ingressclass
+   # Check which controller: nginx, traefik, etc.
+   ```
+
+3. **Update ingress with appropriate annotations** (see above)
+
+4. **Apply changes:**
+   ```bash
+   kubectl apply -f <your-gitea-ingress>.yaml
+   ```
+
+5. **Test push:**
+   ```bash
+   docker push gitea.kazcloud.dev/ryan/turbovault-app:v1.0.0
+   ```
+
+### Alternative: Increase Gitea Service Timeouts
+
+If using a LoadBalancer or NodePort directly:
+
+Edit Gitea's `app.ini`:
+```ini
+[server]
+LFS_MAX_FILE_SIZE = 0
+HTTP_PORT = 3000
+
+[packages]
+ENABLED = true
+CHUNKED_UPLOAD_PATH = /tmp/package-upload
+LIMIT_TOTAL_OWNER_SIZE = -1
+LIMIT_SIZE_CONTAINER = -1
+```
+
+Then restart Gitea pod.
+
+---
+
+## Current Workaround
+
+Using GitHub Container Registry (ghcr.io) for now:
+- Image: `ghcr.io/ryankazokas/turbovault-app:latest`
+- All k8s manifests updated to use ghcr.io
+- GitHub Actions workflow configured
+- Works perfectly, no timeout issues
+
+Once Gitea ingress is fixed, can switch back by updating:
+- k8s/deployment.yaml
+- k8s/migrate-job.yaml
+- .github/workflows/build-and-deploy.yml
+- scripts/update-deployment.sh
+
+---
+
+**Priority:** Low (ghcr.io works fine for now)  
+**Complexity:** Medium (depends on ingress controller setup)  
+**Benefit:** Full control over container registry on your infrastructure