feat(ci): let HCP deploys share an SSM prefix (#149)

Optional task-env replace, source-map upload, health attempts, and a concurrency suffix keep existing callers on the same defaults.

.github/workflows/cd-hcp-fargate.yaml

@@ -18,6 +18,11 @@ name: CD — HCP Fargate
 #         docker-platform: linux/amd64
 #         ship-gate: true
 #
+# apply-task-environment replaces the container env from
+# ${prefix}/task-environment. sentry-project uploads image files before
+# RegisterTaskDefinition. concurrency-suffix splits two deployables that
+# share one SSM prefix. Empty defaults keep the previous behavior.
+#
 # Nothing here creates an HCP run. Terraform owns the cluster, service, ALB,
 # and ignores container_definitions / task_definition.
 
@@ -57,6 +62,36 @@ on:
         type: string
         required: false
         default: "{}"
+      apply-task-environment:
+        description: "Replace container env from SSM ${prefix}/task-environment. GIT_SHA wins."
+        type: boolean
+        required: false
+        default: false
+      sentry-org:
+        description: "Sentry org for BFF source map upload when sentry-project is set"
+        type: string
+        required: false
+        default: "seahaven"
+      sentry-project:
+        description: "Sentry project for BFF source map upload. Empty skips upload."
+        type: string
+        required: false
+        default: ""
+      sentry-container-files:
+        description: "Comma-separated image paths to upload. Required when sentry-project is set."
+        type: string
+        required: false
+        default: ""
+      health-attempts:
+        description: "Number of /api/health polls, 10 seconds apart, before failing"
+        type: number
+        required: false
+        default: 6
+      concurrency-suffix:
+        description: "Optional concurrency group suffix when two deployables share an SSM prefix"
+        type: string
+        required: false
+        default: ""
 
 permissions:
   contents: read
@@ -69,7 +104,7 @@ jobs:
     timeout-minutes: 30
     environment: ${{ inputs.environment }}
     concurrency:
-      group: deploy-${{ inputs.ssm-prefix }}-${{ inputs.environment }}
+      group: ${{ inputs.concurrency-suffix != '' && format('deploy-{0}-{1}-{2}', inputs.ssm-prefix, inputs.concurrency-suffix, inputs.environment) || format('deploy-{0}-{1}', inputs.ssm-prefix, inputs.environment) }}
       cancel-in-progress: false
     env:
       AWS_REGION: us-east-1
@@ -232,6 +267,57 @@ jobs:
             --push \
             .
 
+      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7.0.0
+        if: ${{ inputs.sentry-project != '' }}
+        with:
+          node-version: "24"
+
+      - name: Upload BFF source maps
+        if: ${{ inputs.sentry-project != '' }}
+        env:
+          ECR: ${{ steps.deploy.outputs.ecr }}
+          GIT_SHA: ${{ steps.commit.outputs.sha }}
+          SENTRY_AUTH_TOKEN: ${{ secrets.SENTRY_AUTH_TOKEN }}
+          SENTRY_URL: https://de.sentry.io
+          SENTRY_ORG: ${{ inputs.sentry-org }}
+          SENTRY_PROJECT: ${{ inputs.sentry-project }}
+          SENTRY_CONTAINER_FILES: ${{ inputs.sentry-container-files }}
+        run: |
+          set -euo pipefail
+          if [ -z "${SENTRY_AUTH_TOKEN}" ]; then
+            echo "SENTRY_AUTH_TOKEN is required to upload BFF source maps" >&2
+            exit 1
+          fi
+          if [ -z "${SENTRY_CONTAINER_FILES}" ]; then
+            echo "sentry-container-files is required when sentry-project is set" >&2
+            exit 1
+          fi
+          docker pull "${ECR}:${GIT_SHA}"
+          mkdir -p build/sentry
+          cid="$(docker create "${ECR}:${GIT_SHA}")"
+          cleanup() { docker rm "${cid}" >/dev/null 2>&1 || true; }
+          trap cleanup EXIT
+          IFS=',' read -r -a files <<< "${SENTRY_CONTAINER_FILES}"
+          for path in "${files[@]}"; do
+            path="${path#"${path%%[![:space:]]*}"}"
+            path="${path%"${path##*[![:space:]]}"}"
+            if [ -z "${path}" ]; then
+              echo "sentry-container-files contains an empty path" >&2
+              exit 1
+            fi
+            base="$(basename "${path}")"
+            docker cp "${cid}:${path}" "build/sentry/${base}"
+          done
+          if [ -f build/sentry/server.js ]; then
+            grep -q "${GIT_SHA}" build/sentry/server.js
+            grep -q debugId build/sentry/server.js
+          fi
+          npx --yes @sentry/cli@2 sourcemaps upload \
+            --org "${SENTRY_ORG}" \
+            --project "${SENTRY_PROJECT}" \
+            --release "${GIT_SHA}" \
+            build/sentry
+
       - name: Register task definition and update service
         env:
           CLUSTER: ${{ steps.deploy.outputs.cluster }}
@@ -241,8 +327,19 @@ jobs:
           IMAGE: ${{ steps.deploy.outputs.ecr }}:${{ steps.commit.outputs.sha }}
           GIT_SHA: ${{ steps.commit.outputs.sha }}
           EXTRA_TASK_ENV: ${{ inputs.extra-task-env }}
+          APPLY_TASK_ENVIRONMENT: ${{ inputs.apply-task-environment }}
+          SSM_PREFIX: ${{ inputs.ssm-prefix }}
         run: |
           set -euo pipefail
+          if [ "${APPLY_TASK_ENVIRONMENT}" = "true" ]; then
+            prefix="${SSM_PREFIX%/}"
+            TASK_ENV_JSON="$(aws ssm get-parameter \
+              --name "${prefix}/task-environment" \
+              --with-decryption \
+              --query Parameter.Value \
+              --output text)"
+            export TASK_ENV_JSON
+          fi
           aws ecs describe-task-definition \
             --task-definition "${FAMILY}" \
             --query taskDefinition \
@@ -268,16 +365,25 @@ jobs:
           extra_env = json.loads(extra_raw)
           if not isinstance(extra_env, dict):
               sys.exit("extra-task-env must be a JSON object")
+          apply = os.environ.get("APPLY_TASK_ENVIRONMENT") == "true"
           found = False
           for container in td["containerDefinitions"]:
               if container["name"] != name:
                   continue
               found = True
               container["image"] = image
-              env = {item["name"]: item["value"] for item in container.get("environment", [])}
-              env["GIT_SHA"] = sha
+              if apply:
+                  env_map = json.loads(os.environ["TASK_ENV_JSON"])
+                  if not isinstance(env_map, dict) or not env_map:
+                      sys.exit("task-environment must be a non-empty JSON object")
+                  env = {str(key): str(value) for key, value in env_map.items()}
+                  env.pop("GIT_SHA", None)
+                  container["stopTimeout"] = 60
+              else:
+                  env = {item["name"]: item["value"] for item in container.get("environment", [])}
               for key, value in extra_env.items():
                   env[str(key)] = str(value)
+              env["GIT_SHA"] = sha
               container["environment"] = [{"name": key, "value": value} for key, value in env.items()]
               container.pop("command", None)
           if not found:
@@ -298,6 +404,7 @@ jobs:
           API_URL: ${{ steps.deploy.outputs.api_url }}
           HEALTH_PATH: ${{ inputs.health-path }}
           EXPECTED_SHA: ${{ steps.commit.outputs.sha }}
+          HEALTH_ATTEMPTS: ${{ inputs.health-attempts }}
         run: |
           set -euo pipefail
           path="${HEALTH_PATH}"
@@ -306,7 +413,11 @@ jobs:
             *) path="/${path}" ;;
           esac
           url="${API_URL%/}${path}"
-          for _ in 1 2 3 4 5 6; do
+          if ! [[ "${HEALTH_ATTEMPTS}" =~ ^[1-9][0-9]*$ ]]; then
+            echo "health-attempts must be a positive integer" >&2
+            exit 1
+          fi
+          for _ in $(seq 1 "${HEALTH_ATTEMPTS}"); do
             BODY="$(curl -fsS "${url}" || true)"
             echo "${BODY}"
             if echo "${BODY}" | python3 -c 'import json,os,sys; d=json.load(sys.stdin); sys.exit(0 if d.get("sha")==os.environ["EXPECTED_SHA"] else 1)'; then

.github/workflows/cd-hcp-spa.yaml

@@ -21,6 +21,10 @@ name: CD — HCP SPA
 #         ssm-prefix: /internal-portal/deploy
 #         ship-gate: true
 #
+# concurrency-suffix splits two deployables that share one SSM prefix.
+# verify-companion-api adds cache, asset, and /api/health checks after the
+# index.html hash matches. Empty defaults keep the previous behavior.
+#
 # Nothing here creates an HCP run. Terraform owns the bucket and distribution.
 
 on:
@@ -49,6 +53,16 @@ on:
         type: string
         required: false
         default: ""
+      verify-companion-api:
+        description: "After the index hash matches, check cache headers, hashed assets, and /api/health"
+        type: boolean
+        required: false
+        default: false
+      concurrency-suffix:
+        description: "Optional concurrency group suffix when two deployables share an SSM prefix"
+        type: string
+        required: false
+        default: ""
 
 permissions:
   contents: read
@@ -61,7 +75,7 @@ jobs:
     timeout-minutes: 45
     environment: ${{ inputs.environment }}
     concurrency:
-      group: deploy-${{ inputs.ssm-prefix }}-${{ inputs.environment }}
+      group: ${{ inputs.concurrency-suffix != '' && format('deploy-{0}-{1}-{2}', inputs.ssm-prefix, inputs.concurrency-suffix, inputs.environment) || format('deploy-{0}-{1}', inputs.ssm-prefix, inputs.environment) }}
       cancel-in-progress: false
     env:
       AWS_REGION: us-east-1
@@ -301,3 +315,95 @@ jobs:
           done
           echo "release did not converge: status=${last_status} served_sha256=${last_hash} expected=${EXPECTED_INDEX_SHA256}" >&2
           exit 1
+
+      - name: Verify companion API
+        if: ${{ inputs.verify-companion-api }}
+        env:
+          SITE_URL: ${{ steps.deploy.outputs.site_url }}
+          HEALTH_BUDGET: "20"
+          HEALTH_INTERVAL: "15"
+        run: |
+          set -euo pipefail
+          SITE_URL="${SITE_URL%/}"
+          tmp="$(mktemp -d)"
+          trap 'rm -rf "${tmp}"' EXIT
+
+          curl -fsS --max-time 30 "${SITE_URL}/" -o "${tmp}/index.html" -D "${tmp}/index.headers"
+          curl -fsS --max-time 30 "${SITE_URL}/signin" -o "${tmp}/signin.html"
+          curl -fsS --max-time 30 "${SITE_URL}/help" -o "${tmp}/route.html"
+          if ! grep -qiE 'cache-control:.*no-store' "${tmp}/index.headers"; then
+            echo "FAIL: HTML Cache-Control is missing no-store." >&2
+            exit 1
+          fi
+
+          python3 -c '
+          import re, sys
+          html = open(sys.argv[1], encoding="utf-8").read()
+          seen = []
+          for path in re.findall(r"(?:src|href)=\"(/assets/[^\"]+\.(?:js|css))\"", html):
+              if path not in seen:
+                  seen.append(path)
+                  print(path)
+          ' "${tmp}/index.html" > "${tmp}/asset-paths.txt"
+
+          if [ ! -s "${tmp}/asset-paths.txt" ]; then
+            echo "FAIL: served index.html has no /assets/ URL to check immutable caching." >&2
+            exit 1
+          fi
+          : > "${tmp}/assets.txt"
+          immutable_ok="no"
+          while IFS= read -r asset_path; do
+            curl -fsS --max-time 30 "${SITE_URL}${asset_path}" \
+              -o "${tmp}/asset-body" -D "${tmp}/asset.headers"
+            cat "${tmp}/asset-body" >> "${tmp}/assets.txt"
+            if [[ "${asset_path}" == *.js && "${immutable_ok}" == "no" ]]; then
+              if ! grep -qiE 'cache-control:.*immutable' "${tmp}/asset.headers"; then
+                echo "FAIL: hashed asset is missing Cache-Control immutable." >&2
+                exit 1
+              fi
+              immutable_ok="yes"
+            fi
+          done < "${tmp}/asset-paths.txt"
+          if [ "${immutable_ok}" != "yes" ]; then
+            echo "FAIL: served index.html has no hashed JS asset to check immutable caching." >&2
+            exit 1
+          fi
+          cat "${tmp}/index.html" "${tmp}/assets.txt" > "${tmp}/served.txt"
+          if grep -Eiq 'https?://(localhost|127\.0\.0\.1):[0-9]+' "${tmp}/served.txt"; then
+            echo "FAIL: served assets contain forbidden URL localhost." >&2
+            exit 1
+          fi
+
+          health_code="000"
+          health_sha=""
+          health_attempt=0
+          while [ "${health_attempt}" -lt "${HEALTH_BUDGET}" ]; do
+            health_attempt=$((health_attempt + 1))
+            health_code="$(curl -sS --max-time 30 -o "${tmp}/health.json" -w '%{http_code}' "${SITE_URL}/api/health" || echo "000")"
+            echo "health poll ${health_attempt}/${HEALTH_BUDGET}: GET /api/health http=${health_code}"
+            if [ "${health_code}" = "200" ]; then
+              health_sha="$(python3 -c 'import json,sys
+          try:
+              print(json.load(open(sys.argv[1], encoding="utf-8")).get("sha") or "")
+          except Exception:
+              print("")
+          ' "${tmp}/health.json")"
+              if [ -n "${health_sha}" ] && [ "${health_sha}" != "bootstrap" ]; then
+                break
+              fi
+              echo "health poll ${health_attempt}/${HEALTH_BUDGET}: sha=${health_sha:-missing} (waiting for Deploy API)"
+            fi
+            if [ "${health_attempt}" -lt "${HEALTH_BUDGET}" ]; then
+              sleep "${HEALTH_INTERVAL}"
+            fi
+          done
+          if [ "${health_code}" != "200" ]; then
+            echo "FAIL: GET /api/health returned HTTP ${health_code} after ${HEALTH_BUDGET} polls." >&2
+            exit 1
+          fi
+          if [ -z "${health_sha}" ] || [ "${health_sha}" = "bootstrap" ]; then
+            echo "FAIL: GET /api/health is still the bootstrap stub after ${HEALTH_BUDGET} polls." >&2
+            exit 1
+          fi
+          python3 -c 'import json,sys; body=json.load(open(sys.argv[1], encoding="utf-8")); raise SystemExit(0 if body.get("stage") and body.get("sha") else 1)' "${tmp}/health.json"
+          echo "PASS: companion API smoke checks passed."