From 56452302f2dac6c66c34c9b90d9b50a23e6ea7a3 Mon Sep 17 00:00:00 2001
From: Sourav Chandhok <114119117+souravC01@users.noreply.github.com>
Date: Sat, 12 Sep 2026 03:57:30 -0400
Subject: [PATCH] ci: deploy backend to Cloud Run with keyless authentication

---
 .dockerignore                   |   1 +
 .github/workflows/ci.yml        | 124 ++++++++++++++++++++++++++++----
 .gitignore                      |   1 +
 docs/cloud-run-cicd.md          |  86 ++++++++++++++++++++++
 scripts/setup-cloud-run-cicd.sh |  56 +++++++++++++++
 5 files changed, 255 insertions(+), 13 deletions(-)
 create mode 100644 docs/cloud-run-cicd.md
 create mode 100644 scripts/setup-cloud-run-cicd.sh

diff --git a/.dockerignore b/.dockerignore
index 5aeaaae..5762b5f 100644
--- a/.dockerignore
+++ b/.dockerignore
@@ -21,3 +21,4 @@ scratch
 .DS_Store
 Thumbs.db
 *.log
+gha-creds-*.json
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 5fa4e61..4cfed15 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -15,9 +15,49 @@ name: CI/CD Pipeline
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: true
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+  queue: ${{ github.ref == 'refs/heads/main' && 'max' || 'single' }}
+
+permissions:
+  contents: read
 
 jobs:
+  backend-changes:
+    name: Detect Backend Changes
+    if: github.event_name == 'push' && github.ref == 'refs/heads/main'
+    runs-on: ubuntu-latest
+    outputs:
+      changed: ${{ steps.changes.outputs.changed }}
+    steps:
+      - name: Checkout repository history
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Check backend deployment paths
+        id: changes
+        env:
+          BEFORE_SHA: ${{ github.event.before }}
+          AFTER_SHA: ${{ github.sha }}
+        shell: bash
+        run: |
+          set -euo pipefail
+          if [[ "$BEFORE_SHA" == "0000000000000000000000000000000000000000" ]]; then
+            BEFORE_SHA=$(git hash-object -t tree /dev/null)
+          elif ! git cat-file -e "${BEFORE_SHA}^{commit}" 2>/dev/null; then
+            git fetch --no-tags origin "$BEFORE_SHA"
+          fi
+          if git diff --quiet "$BEFORE_SHA" "$AFTER_SHA" -- \
+            backend/ requirements.txt requirements/ Dockerfile.cloudrun .github/workflows/ci.yml; then
+            echo "changed=false" >> "$GITHUB_OUTPUT"
+          else
+            result=$?
+            if [[ "$result" -ne 1 ]]; then
+              exit "$result"
+            fi
+            echo "changed=true" >> "$GITHUB_OUTPUT"
+          fi
+
   backend-ci:
     name: Backend Tests & Validation
     runs-on: ubuntu-latest
@@ -115,6 +155,14 @@ jobs:
           push: false
           tags: jobhelperguru:ci-test
 
+      - name: Build Cloud Run container image
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          file: Dockerfile.cloudrun
+          push: false
+          tags: jobhelperguru-backend:ci-test
+
   ci-success:
     name: CI Pipeline Passed
     needs: [backend-ci, frontend-ci, docker-ci]
@@ -123,18 +171,68 @@ jobs:
       - name: Validate All Required Checks Passed
         run: echo "All backend, frontend, and Docker validation checks passed successfully."
 
-  deploy-render:
-    name: Trigger Render Deployment
-    needs: ci-success
-    if: github.ref == 'refs/heads/main' && github.event_name == 'push'
+  deploy-cloud-run:
+    name: Deploy Cloud Run Backend
+    needs: [ci-success, backend-changes]
+    if: >-
+      github.event_name == 'push' && github.ref == 'refs/heads/main' &&
+      needs.backend-changes.outputs.changed == 'true'
     runs-on: ubuntu-latest
+    timeout-minutes: 30
+    permissions:
+      contents: read
+      id-token: write
+    env:
+      IMAGE: us-east5-docker.pkg.dev/jobhelperguru/jobhelperguru/jobhelperguru-backend:${{ github.sha }}
     steps:
-      - name: Trigger Render Deploy Hook
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      # Build before requesting short-lived credentials.
+      - name: Build backend image
+        run: docker build --file Dockerfile.cloudrun --tag "$IMAGE" .
+
+      - name: Authenticate to Google Cloud
+        uses: google-github-actions/auth@7c6bc770dae815cd3e89ee6cdf493a5fab2cc093 # v3
+        with:
+          project_id: jobhelperguru
+          workload_identity_provider: projects/999060759573/locations/global/workloadIdentityPools/jobhelperguru-github/providers/github-main
+          service_account: jobhelperguru-github-deployer@jobhelperguru.iam.gserviceaccount.com
+
+      - name: Set up Google Cloud CLI
+        uses: google-github-actions/setup-gcloud@aa5489c8933f4cc7a4f7d45035b3b1440c9c10db # v3
+
+      - name: Push backend image
         run: |
-          if [ -z "${{ secrets.RENDER_DEPLOY_HOOK_URL }}" ]; then
-            echo "::warning ::RENDER_DEPLOY_HOOK_URL secret is not configured in repository secrets. Skipping deploy hook trigger."
-          else
-            echo "Triggering Render deployment..."
-            curl -f -s -S -X POST "${{ secrets.RENDER_DEPLOY_HOOK_URL }}"
-            echo "Render deploy hook triggered successfully!"
-          fi
+          gcloud auth configure-docker us-east5-docker.pkg.dev --quiet
+          docker push "$IMAGE"
+
+      - name: Update Cloud Run image
+        run: |
+          gcloud run services update jobhelperguru-backend \
+            --project=jobhelperguru \
+            --region=us-east5 \
+            --image="$IMAGE" \
+            --quiet
+
+      - name: Verify production health through Vercel
+        shell: python
+        run: |
+          import json
+          import time
+          import urllib.request
+
+          url = "https://jobhelperguru.vercel.app/api/health"
+          for attempt in range(12):
+              try:
+                  with urllib.request.urlopen(url, timeout=20) as response:
+                      health = json.load(response)
+                  assert health.get("status") == "ok", "Backend is not healthy"
+                  assert health.get("database") == "postgresql", "Unexpected database"
+                  assert health.get("object_storage") == "cloudflare_r2", "Unexpected object storage"
+                  print("Production health verified: ok, postgresql, cloudflare_r2")
+                  break
+              except Exception:
+                  if attempt == 11:
+                      raise
+                  time.sleep(10)
diff --git a/.gitignore b/.gitignore
index 15697ce..e2ec16f 100644
--- a/.gitignore
+++ b/.gitignore
@@ -10,3 +10,4 @@ dist/
 .DS_Store
 scratch/
 docs/superpowers/
+gha-creds-*.json
diff --git a/docs/cloud-run-cicd.md b/docs/cloud-run-cicd.md
new file mode 100644
index 0000000..88fb700
--- /dev/null
+++ b/docs/cloud-run-cicd.md
@@ -0,0 +1,86 @@
+# Backend continuous deployment
+
+The frontend stays on Vercel. The backend stays on the existing Cloud Run service
+`jobhelperguru-backend` in project `jobhelperguru`, region `us-east5`.
+The Render legacy redirect and its `render-legacy-redirect` branch are independent.
+The main CI workflow no longer calls the Render deploy hook.
+
+## One-time Google Cloud setup
+
+Run `bash scripts/setup-cloud-run-cicd.sh` from an administrator's authenticated
+Google Cloud Shell after reviewing the script. It checks the existing project,
+Artifact Registry repository, and runtime service account before creating the
+dedicated federation resources. It stops if the dedicated account or pool already
+exists; inspect partial setup before resuming individual commands.
+
+The deployer is `jobhelperguru-github-deployer@jobhelperguru.iam.gserviceaccount.com`.
+Its resource-scoped grants are:
+
+| Role | Resource |
+| --- | --- |
+| Artifact Registry Writer | `jobhelperguru` repository in `us-east5` |
+| Cloud Run Developer | Existing `jobhelperguru-backend` service |
+| Service Account User | Existing `jobhelperguru-runner` runtime account |
+
+The dedicated pool `jobhelperguru-github` and provider `github-main` accept only
+GitHub OIDC tokens matching repository ID `1355476075`, owner ID `114119117`,
+repository `souravC01/JobHelperGuru`, branch `refs/heads/main`, event `push`, and
+workflow `souravC01/JobHelperGuru/.github/workflows/ci.yml@refs/heads/main`.
+The deployer impersonation binding additionally requires the exact main-branch
+OIDC subject. Pull requests, forks, other branches, manual dispatches, and other
+workflow files cannot use this provider. No service-account JSON keys or new
+GitHub secrets are needed. The provider identifier and account email in the
+workflow are public configuration.
+
+## Deployment behavior
+
+Existing backend tests, frontend tests/build, Docker validation, and `ci-success`
+remain. Docker validation also builds `Dockerfile.cloudrun` before a PR is merged.
+
+On pushes to `main`, native Git compares the push's before and after commits for
+changes to `backend/`, `requirements.txt`, `requirements/`, `Dockerfile.cloudrun`,
+or `.github/workflows/ci.yml`. This covers all commits in a push, including deleted
+files. An initial branch push compares against an empty tree. An unavailable
+previous commit is fetched; a comparison failure fails the detector.
+
+After CI succeeds and a relevant change is found, the workflow builds the backend,
+requests short-lived credentials using `google-github-actions/auth`, pushes
+`us-east5-docker.pkg.dev/jobhelperguru/jobhelperguru/jobhelperguru-backend:<commit-sha>`,
+and runs `gcloud run services update` with only the image, project, and region.
+Building precedes authentication so a long build does not consume the credentials'
+short lifetime. Credential files are excluded from Git and Docker build contexts.
+
+The image update preserves environment variables, secret references, scaling,
+public access, traffic policy, and the runtime account. Production currently routes
+100 percent of traffic to the latest revision. Main runs are not cancelled when a
+new push arrives, so an in-progress deployment can finish. Main runs use GitHub's
+maximum queue of 100 pending runs so a frontend-only push cannot replace a pending
+backend deployment. Other branches retain cancellation of superseded CI runs.
+
+The final step retries `https://jobhelperguru.vercel.app/api/health` and requires
+`status=ok`, `database=postgresql`, and `object_storage=cloudflare_r2`. If deployment
+or health verification fails, the workflow fails and needs investigation; it does
+not automatically roll back.
+
+## Verification and recovery
+
+Before merging, verify workflow syntax and all PR CI checks. After the first main
+run, verify its successful deployment step and Cloud Run's latest ready revision
+and image SHA, then compare the service configuration with the pre-deployment
+snapshot. Health alone does not prove which image is serving traffic.
+
+To restore a previously verified image, an authorized operator can run:
+
+```bash
+gcloud run services update jobhelperguru-backend \
+  --project=jobhelperguru --region=us-east5 \
+  --image=us-east5-docker.pkg.dev/jobhelperguru/jobhelperguru/jobhelperguru-backend:PREVIOUS_SHA
+```
+
+Use an existing verified SHA from Artifact Registry. To stop future backend
+deployments, disable this federation provider or revert the deployment job. Do not
+restore the Render deploy hook. Follow the same Vercel health check after recovery.
+
+References: [Google federation setup](https://cloud.google.com/iam/docs/workload-identity-federation-with-deployment-pipelines),
+[Cloud Run deployment roles](https://cloud.google.com/run/docs/deploying), and
+[Google authentication action](https://github.com/google-github-actions/auth).
diff --git a/scripts/setup-cloud-run-cicd.sh b/scripts/setup-cloud-run-cicd.sh
new file mode 100644
index 0000000..c980722
--- /dev/null
+++ b/scripts/setup-cloud-run-cicd.sh
@@ -0,0 +1,56 @@
+#!/usr/bin/env bash
+# One-time bootstrap. Run as a project administrator in Google Cloud Shell.
+set -euo pipefail
+
+project=jobhelperguru
+project_number=999060759573
+region=us-east5
+pool=jobhelperguru-github
+provider=github-main
+deployer=jobhelperguru-github-deployer@jobhelperguru.iam.gserviceaccount.com
+runtime=jobhelperguru-runner@jobhelperguru.iam.gserviceaccount.com
+
+actual_number=$(gcloud projects describe "$project" --format='value(projectNumber)')
+[[ "$actual_number" == "$project_number" ]] || { echo 'Unexpected project number'; exit 1; }
+actual_runtime=$(gcloud run services describe jobhelperguru-backend \
+  --project="$project" --region="$region" --format='value(spec.template.spec.serviceAccountName)')
+[[ "$actual_runtime" == "$runtime" ]] || { echo 'Unexpected runtime service account'; exit 1; }
+gcloud artifacts repositories describe jobhelperguru --project="$project" --location="$region" --format='value(name)'
+
+gcloud services enable iam.googleapis.com iamcredentials.googleapis.com sts.googleapis.com --project="$project"
+
+# Creation fails closed if these dedicated resources already exist. Inspect any
+# partial setup before resuming instead of replacing an existing trust policy.
+gcloud iam service-accounts create jobhelperguru-github-deployer \
+  --project="$project" --display-name='JobHelperGuru GitHub deployer'
+gcloud iam workload-identity-pools create "$pool" \
+  --project="$project" --location=global --display-name='JobHelperGuru GitHub'
+
+# Numeric IDs prevent a renamed or recreated repository or owner from inheriting
+# access. Also require the exact repository, branch, event, and workflow file.
+condition="assertion.repository_id == '1355476075' && assertion.repository_owner_id == '114119117' && assertion.repository == 'souravC01/JobHelperGuru' && assertion.ref == 'refs/heads/main' && assertion.event_name == 'push' && assertion.workflow_ref == 'souravC01/JobHelperGuru/.github/workflows/ci.yml@refs/heads/main' && assertion.sub == 'repo:souravC01/JobHelperGuru:ref:refs/heads/main'"
+gcloud iam workload-identity-pools providers create-oidc "$provider" \
+  --project="$project" --location=global --workload-identity-pool="$pool" \
+  --display-name='GitHub main CI workflow' \
+  --issuer-uri='https://token.actions.githubusercontent.com' \
+  --attribute-mapping='google.subject=assertion.sub,attribute.repository_id=assertion.repository_id,attribute.repository_owner_id=assertion.repository_owner_id' \
+  --attribute-condition="$condition"
+
+gcloud iam service-accounts add-iam-policy-binding "$deployer" \
+  --project="$project" --role=roles/iam.workloadIdentityUser \
+  --member="principal://iam.googleapis.com/projects/$project_number/locations/global/workloadIdentityPools/$pool/subject/repo:souravC01/JobHelperGuru:ref:refs/heads/main" \
+  --condition=None
+gcloud artifacts repositories add-iam-policy-binding jobhelperguru \
+  --project="$project" --location="$region" \
+  --role=roles/artifactregistry.writer --member="serviceAccount:$deployer" --condition=None
+gcloud run services add-iam-policy-binding jobhelperguru-backend \
+  --project="$project" --region="$region" \
+  --role=roles/run.developer --member="serviceAccount:$deployer" --condition=None
+gcloud iam service-accounts add-iam-policy-binding "$runtime" \
+  --project="$project" --role=roles/iam.serviceAccountUser \
+  --member="serviceAccount:$deployer" --condition=None
+
+echo 'Keyless deployment identity configured. No service-account keys were created.'
+gcloud iam workload-identity-pools providers describe "$provider" \
+  --project="$project" --location=global --workload-identity-pool="$pool" \
+  --format='yaml(name,state,attributeMapping,attributeCondition,oidc)'