chore(deploy): automate production rollout

.gitea/workflows/build-production-docker.yml

@@ -24,6 +24,7 @@ jobs:
           test -n "${{ secrets.AWS_REGION }}"
           test -n "${{ secrets.AWS_ACCOUNT_ID }}"
           test -n "${{ secrets.ECR_REPOSITORY }}"
+          test -n "${{ secrets.DEPLOY_SSH_KEY }}"
 
       - name: Build backend production image
         run: |
@@ -33,7 +34,6 @@ jobs:
             --load \
             --file backend/Dockerfile \
             --tag ${IMAGE_NAME}:prod-${GITHUB_SHA} \
-            --tag ${IMAGE_NAME}:prod-latest \
             .
 
       - name: Check image size budget
@@ -88,7 +88,81 @@ jobs:
           aws ecr get-login-password --region "${AWS_REGION}" \
             | docker login --username AWS --password-stdin "${ecr_registry}"
 
-          docker tag "${IMAGE_NAME}:prod-${GITHUB_SHA}" "${ecr_image}:${GITHUB_SHA}"
           docker tag "${IMAGE_NAME}:prod-${GITHUB_SHA}" "${ecr_image}:latest"
-          docker push "${ecr_image}:${GITHUB_SHA}"
+
+          image_ids="$(aws ecr list-images \
+            --region "${AWS_REGION}" \
+            --repository-name "${ECR_REPOSITORY}" \
+            --query 'imageIds[*]' \
+            --output json)"
+          if [ "${image_ids}" != "[]" ]; then
+            aws ecr batch-delete-image \
+              --region "${AWS_REGION}" \
+              --repository-name "${ECR_REPOSITORY}" \
+              --image-ids "${image_ids}" >/dev/null
+          fi
+
           docker push "${ecr_image}:latest"
+
+  deploy-production:
+    needs: build-backend-image
+    runs-on: wsl2-docker-host
+    steps:
+      - name: Validate deploy configuration
+        run: |
+          set -euo pipefail
+          test -n "${{ secrets.DEPLOY_SSH_KEY }}"
+          test -n "${{ secrets.DEPLOY_HOST }}"
+          test -n "${{ secrets.DEPLOY_USER }}"
+          test -n "${{ secrets.AWS_ACCESS_KEY_ID }}"
+          test -n "${{ secrets.AWS_SECRET_ACCESS_KEY }}"
+          test -n "${{ secrets.AWS_REGION }}"
+
+      - name: Deploy production server
+        env:
+          AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
+          AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
+          AWS_DEFAULT_REGION: ${{ secrets.AWS_REGION }}
+          AWS_REGION: ${{ secrets.AWS_REGION }}
+          DEPLOY_HOST: ${{ secrets.DEPLOY_HOST }}
+          DEPLOY_USER: ${{ secrets.DEPLOY_USER }}
+        run: |
+          set -euo pipefail
+
+          install -m 700 -d ~/.ssh
+          printf '%s\n' '${{ secrets.DEPLOY_SSH_KEY }}' > ~/.ssh/eryao_deploy_key
+          chmod 600 ~/.ssh/eryao_deploy_key
+          ssh-keyscan -H "${DEPLOY_HOST}" >> ~/.ssh/known_hosts
+
+          ssh -i ~/.ssh/eryao_deploy_key \
+            -o IdentitiesOnly=yes \
+            "${DEPLOY_USER}@${DEPLOY_HOST}" \
+            "AWS_ACCESS_KEY_ID='${AWS_ACCESS_KEY_ID}' AWS_SECRET_ACCESS_KEY='${AWS_SECRET_ACCESS_KEY}' AWS_DEFAULT_REGION='${AWS_REGION}' AWS_REGION='${AWS_REGION}' bash -se" <<'REMOTE'
+          set -euo pipefail
+
+          cd ~/deploy
+          set -a
+          . ./.env
+          set +a
+
+          ecr_registry="${AWS_ACCOUNT_ID}.dkr.ecr.${AWS_REGION}.amazonaws.com"
+          aws ecr get-login-password --region "${AWS_REGION}" \
+            | sudo docker login --username AWS --password-stdin "${ecr_registry}"
+
+          sudo docker compose --env-file ./.env -f docker-compose.prod.yml --profile workers pull
+          sudo docker compose --env-file ./.env -f docker-compose.prod.yml --profile workers up -d --remove-orphans
+
+          for attempt in $(seq 1 12); do
+            if curl -fsS "http://127.0.0.1:${ERYAO_WEB__PORT:-5775}/health"; then
+              break
+            fi
+            if [ "${attempt}" -eq 12 ]; then
+              sudo docker compose --env-file ./.env -f docker-compose.prod.yml --profile workers ps
+              sudo docker logs --tail 200 eryao-prod-backend || true
+              exit 1
+            fi
+            sleep 5
+          done
+
+          sudo docker image prune -af --filter "until=168h"
+          REMOTE

.gitignore

@@ -137,6 +137,11 @@ ENV/
 env.bak/
 venv.bak/
 
+# Local deployment secrets
+*.pem
+deploy_eryao_ci
+deploy_eryao_ci.pub
+
 # Spyder project settings
 .spyderproject
 .spyproject

deploy/README.md

@@ -151,7 +151,30 @@ docker logs -f eryao-prod-redis
 
 ## 更新版本
 
-CI 推送新镜像到 ECR 后，在生产机器执行：
+当前 CI/CD 会在 `main` 分支构建后自动部署到生产机器：
+
+- 推送前清空 ECR 仓库旧镜像，只保留新推送的 `latest`。
+- 通过 `DEPLOY_SSH_KEY` 登录生产机器。
+- 在生产机器执行 ECR 登录、`docker compose pull`、`docker compose up -d`。
+- 健康检查通过后清理 7 天前未使用的本地 Docker 镜像。
+
+Gitea Secrets 需要包含：
+
+```text
+AWS_ACCESS_KEY_ID
+AWS_SECRET_ACCESS_KEY
+AWS_REGION
+AWS_ACCOUNT_ID
+ECR_REPOSITORY
+DEPLOY_SSH_KEY
+DEPLOY_HOST
+DEPLOY_USER
+```
+
+`DEPLOY_SSH_KEY` 是已加入生产机器 `ubuntu` 用户 `~/.ssh/authorized_keys` 的部署专用私钥。
+当前生产机器对应：`DEPLOY_HOST=18.218.38.213`，`DEPLOY_USER=ubuntu`。
+
+如需手动更新，在生产机器执行：
 
 ```bash
 cd deploy