feat: cost-guard komplett — dev-budsjett, Pub/Sub auto-teardown, webhook-alert, cost-check i CI/CD

2026-05-23 23:22:39 +02:00 · 2026-05-23 23:22:39 +02:00 · aff36f2f35
commit aff36f2f35
parent 4502ae09fd
4 changed files with 295 additions and 14 deletions
--- a/cloudbuild.yaml
+++ b/cloudbuild.yaml
@ -1,7 +1,20 @@
-# cloudbuild.yaml — Auto-build and deploy osvauco-agent on push to main
+# cloudbuild.yaml — Auto-build, cost-check og deploy osvauco-agent på push til main
 steps:
-  # Step 1: Build Docker image
+  # Steg 1: Kost-sjekk før deploy
+  - name: 'gcr.io/google.com/cloudsdktool/cloud-sdk'
+    id: cost-check
+    entrypoint: bash
+    args:
+      - '-c'
+      - |
+        echo "=== Cost Check ==="
+        bash infrastructure/09-cost-check.sh || true
+        echo "=== Cost Check ferdig ==="
+
+  # Steg 2: Bygg Docker-image
  - name: 'gcr.io/cloud-builders/docker'
+    id: build
+    waitFor: ['cost-check']
    args:
      - 'build'
      - '-t'
@ -10,15 +23,19 @@ steps:
      - '${_REGION}-docker.pkg.dev/${PROJECT_ID}/${_ARTIFACT_REPO}/osvauco-agent:latest'
      - 'agents/core-logic'

-  # Step 2: Push image
+  # Steg 3: Push image
  - name: 'gcr.io/cloud-builders/docker'
+    id: push
+    waitFor: ['build']
    args:
      - 'push'
      - '--all-tags'
      - '${_REGION}-docker.pkg.dev/${PROJECT_ID}/${_ARTIFACT_REPO}/osvauco-agent'

-  # Step 3: Deploy to Cloud Run
+  # Steg 4: Deploy til Cloud Run
  - name: 'gcr.io/google.com/cloudsdktool/cloud-sdk'
+    id: deploy
+    waitFor: ['push']
    entrypoint: gcloud
    args:
      - 'run'
@ -37,6 +54,27 @@ steps:
      - '--max-instances=3'
      - '--quiet'

+  # Steg 5: Deploy-notifikasjon til webhook/chat-app
+  - name: 'gcr.io/google.com/cloudsdktool/cloud-sdk'
+    id: notify
+    waitFor: ['deploy']
+    entrypoint: bash
+    args:
+      - '-c'
+      - |
+        WEBHOOK_URL=$(gcloud secrets versions access latest \
+          --secret=webhook-url \
+          --project=${PROJECT_ID} 2>/dev/null || echo "")
+        if [[ -n "${WEBHOOK_URL}" ]]; then
+          MSG="✅ *OSVauco Deploy fullført*\nCommit: \`$COMMIT_SHA\`\nService: \`${_CLOUD_RUN_SERVICE}\`\nRegion: \`${_REGION}\`"
+          curl -s -X POST "${WEBHOOK_URL}" \
+            -H 'Content-Type: application/json' \
+            -d "{\"text\": \"$MSG\"}" || true
+          echo "✓ Deploy-notifikasjon sendt"
+        else
+          echo "⚠️ Ingen webhook-url i Secret Manager — hopper over varsling"
+        fi
+
 substitutions:
  _REGION: us-central1
  _ARTIFACT_REPO: osvauco-repo
--- a/infrastructure/01-setupenv.sh
+++ b/infrastructure/01-setupenv.sh
@ -1,5 +1,5 @@
 #!/bin/bash
-# 01-setupenv.sh — Enable APIs, create staging bucket with lifecycle, SA, billing budget
+# 01-setupenv.sh — Enable APIs, create staging bucket with lifecycle, SA, billing budgets
 # Run in: local VS Code terminal OR Cloud Shell
 # Source .env before running: source .env

@ -11,7 +11,9 @@ set -euo pipefail
 : "${AGENT_SA:?Set AGENT_SA in .env}"

 BUCKET_NAME="gs://${PROJECT_ID}-agent-staging"
-BUDGET_NAME="OSVauco-Agent-Budget-500USD"
+BUDGET_PROD_NAME="OSVauco-Agent-Budget-500USD"
+BUDGET_DEV_NAME="OSVauco-Dev-Budget-75USD"
+PUBSUB_TOPIC="billing-alert-auto-teardown"

 echo "=== 01: Environment Setup for ${PROJECT_ID} ==="

@ -29,10 +31,13 @@ gcloud services enable \
  run.googleapis.com \
  artifactregistry.googleapis.com \
  cloudbuild.googleapis.com \
+  cloudbuildv2.googleapis.com \
  secretmanager.googleapis.com \
  monitoring.googleapis.com \
  logging.googleapis.com \
  cloudtrace.googleapis.com \
+  pubsub.googleapis.com \
+  cloudfunctions.googleapis.com \
  --project="$PROJECT_ID" --quiet
 echo "✓ APIs enabled."

@ -73,24 +78,55 @@ for ROLE in \
 done
 echo "✓ IAM bindings configured."

-# 5. Billing budget alert (idempotent check by display name)
-EXISTING=$(gcloud billing budgets list \
+# 5. Pub/Sub topic for billing auto-teardown (idempotent)
+if ! gcloud pubsub topics describe "${PUBSUB_TOPIC}" --project="${PROJECT_ID}" >/dev/null 2>&1; then
+  gcloud pubsub topics create "${PUBSUB_TOPIC}" --project="${PROJECT_ID}" --quiet
+  echo "✓ Pub/Sub topic created: ${PUBSUB_TOPIC}"
+else
+  echo "✓ Pub/Sub topic already exists: ${PUBSUB_TOPIC}"
+fi
+
+# 6. Prod billing budget — $500, varsler på 50% / 80% / 100% + Pub/Sub (idempotent)
+EXISTING_PROD=$(gcloud billing budgets list \
  --billing-account="${BILLING_ACCOUNT_ID}" \
-  --filter="displayName=${BUDGET_NAME}" \
+  --filter="displayName=${BUDGET_PROD_NAME}" \
  --format="value(name)" 2>/dev/null | head -1 || true)

-if [[ -z "${EXISTING}" ]]; then
+if [[ -z "${EXISTING_PROD}" ]]; then
  gcloud billing budgets create \
    --billing-account="${BILLING_ACCOUNT_ID}" \
-    --display-name="${BUDGET_NAME}" \
+    --display-name="${BUDGET_PROD_NAME}" \
    --budget-amount=500USD \
    --threshold-rule=percent=0.5 \
    --threshold-rule=percent=0.8 \
-    --threshold-rule=percent=1.0
-  echo "✓ Budget alert created"
+    --threshold-rule=percent=1.0 \
+    --notifications-rule-pubsub-topic="projects/${PROJECT_ID}/topics/${PUBSUB_TOPIC}"
+  echo "✓ Prod budget alert created ($500, Pub/Sub koblet)"
 else
-  echo "✓ Budget alert already exists"
+  echo "✓ Prod budget alert already exists"
+fi
+
+# 7. Dev billing budget — $75, varsler på 70% / 90% + Pub/Sub (idempotent)
+EXISTING_DEV=$(gcloud billing budgets list \
+  --billing-account="${BILLING_ACCOUNT_ID}" \
+  --filter="displayName=${BUDGET_DEV_NAME}" \
+  --format="value(name)" 2>/dev/null | head -1 || true)
+
+if [[ -z "${EXISTING_DEV}" ]]; then
+  gcloud billing budgets create \
+    --billing-account="${BILLING_ACCOUNT_ID}" \
+    --display-name="${BUDGET_DEV_NAME}" \
+    --budget-amount=75USD \
+    --threshold-rule=percent=0.7 \
+    --threshold-rule=percent=0.9 \
+    --threshold-rule=percent=1.0 \
+    --notifications-rule-pubsub-topic="projects/${PROJECT_ID}/topics/${PUBSUB_TOPIC}"
+  echo "✓ Dev budget alert created ($75, Pub/Sub koblet)"
+else
+  echo "✓ Dev budget alert already exists"
 fi

 echo ""
 echo "=== 01: Environment setup COMPLETE ==="
+echo "  Kjør nå: bash infrastructure/10-cost-guard.sh"
+echo "  for å aktivere auto-teardown Cloud Function"
--- a/infrastructure/10-cost-guard.sh
+++ b/infrastructure/10-cost-guard.sh
@ -0,0 +1,156 @@
+#!/usr/bin/env bash
+# 10-cost-guard.sh — Deploy Cloud Function som lytter på billing Pub/Sub
+# og utfører auto-teardown + sender webhook-alert til chat-app
+# ved 100% av budsjett.
+#
+# Forutsetter:
+#   - Pub/Sub topic 'billing-alert-auto-teardown' finnes (01-setupenv.sh)
+#   - Secret Manager secret 'webhook-url' inneholder chat-webhook URL
+#     gcloud secrets create webhook-url --data-file=<(echo -n "https://DIN-WEBHOOK-URL")
+# Source .env før kjøring: source .env
+set -euo pipefail
+
+: "${PROJECT_ID:?Set PROJECT_ID}"
+: "${REGION:?Set REGION}"
+: "${CLOUD_RUN_SERVICE:?Set CLOUD_RUN_SERVICE}"
+
+FUNCTION_NAME="billing-auto-teardown"
+PUBSUB_TOPIC="billing-alert-auto-teardown"
+FUNC_DIR="/tmp/cost-guard-fn"
+
+echo "=== 10: Cost Guard — Cloud Function deployment ==="
+
+bash "$(dirname "$0")/00-authcheck.sh"
+
+gcloud services enable cloudfunctions.googleapis.com pubsub.googleapis.com --project="${PROJECT_ID}" --quiet
+
+# --- Skriv Cloud Function kildekode ---
+mkdir -p "${FUNC_DIR}"
+
+cat > "${FUNC_DIR}/main.py" << 'PYEOF'
+import base64
+import json
+import os
+import subprocess
+import urllib.request
+
+import functions_framework
+from google.cloud import secretmanager
+
+
+def get_webhook_url(project_id: str) -> str | None:
+    try:
+        client = secretmanager.SecretManagerServiceClient()
+        name = f"projects/{project_id}/secrets/webhook-url/versions/latest"
+        response = client.access_secret_version(request={"name": name})
+        return response.payload.data.decode("UTF-8")
+    except Exception:
+        return None
+
+
+def send_webhook(url: str, message: str) -> None:
+    payload = json.dumps({"text": message}).encode()
+    req = urllib.request.Request(
+        url,
+        data=payload,
+        headers={"Content-Type": "application/json"},
+        method="POST",
+    )
+    with urllib.request.urlopen(req, timeout=10):
+        pass
+
+
+@functions_framework.cloud_event
+def billing_alert(cloud_event):
+    project_id = os.environ.get("PROJECT_ID", "")
+    service = os.environ.get("CLOUD_RUN_SERVICE", "")
+    region = os.environ.get("REGION", "")
+
+    # Dekod Pub/Sub-melding
+    data = base64.b64decode(cloud_event.data["message"]["data"]).decode()
+    alert = json.loads(data)
+
+    budget_amount = alert.get("budgetAmount", "?")
+    cost_amount = alert.get("costAmount", "?")
+    alert_threshold = alert.get("alertThresholdExceeded", 0)
+
+    print(f"Billing alert: {cost_amount} / {budget_amount} ({alert_threshold*100:.0f}%)")
+
+    webhook_url = get_webhook_url(project_id)
+    base_msg = (
+        f"🚨 *OSVauco Cost Alert*\n"
+        f"Prosjekt: `{project_id}`\n"
+        f"Forbruk: `{cost_amount}` av budsjett `{budget_amount}`\n"
+        f"Terskel: `{alert_threshold*100:.0f}%`"
+    )
+
+    # Kjør auto-teardown kun ved 100% terskel
+    if alert_threshold >= 1.0:
+        print("100% terskel nådd — starter auto-teardown")
+        try:
+            subprocess.run(
+                [
+                    "gcloud", "run", "services", "delete", service,
+                    f"--region={region}",
+                    f"--project={project_id}",
+                    "--quiet",
+                ],
+                check=True,
+                capture_output=True,
+            )
+            teardown_msg = base_msg + "\n\nℹ️ Auto-teardown utført: Cloud Run stoppet."
+        except subprocess.CalledProcessError as e:
+            teardown_msg = base_msg + f"\n\n⚠️ Auto-teardown FEILET: {e.stderr.decode()}"
+        msg = teardown_msg
+    else:
+        msg = base_msg + "\n\n⚠️ Ingen tiltak nå — men pass på forbruket."
+
+    if webhook_url:
+        try:
+            send_webhook(webhook_url, msg)
+            print("Webhook-alert sendt")
+        except Exception as ex:
+            print(f"Webhook feilet: {ex}")
+    else:
+        print("Ingen webhook-url i Secret Manager — hopper over varsling")
+PYEOF
+
+cat > "${FUNC_DIR}/requirements.txt" << 'EOF'
+functions-framework==3.*
+google-cloud-secret-manager>=2.0.0
+EOF
+
+echo "✓ Cloud Function kildekode skrevet til ${FUNC_DIR}"
+
+# --- Deploy Cloud Function ---
+gcloud functions deploy "${FUNCTION_NAME}" \
+  --gen2 \
+  --runtime=python312 \
+  --region="${REGION}" \
+  --source="${FUNC_DIR}" \
+  --entry-point=billing_alert \
+  --trigger-topic="${PUBSUB_TOPIC}" \
+  --project="${PROJECT_ID}" \
+  --set-env-vars="PROJECT_ID=${PROJECT_ID},CLOUD_RUN_SERVICE=${CLOUD_RUN_SERVICE},REGION=${REGION}" \
+  --service-account="${AGENT_SA:-$(gcloud config get-value account)}" \
+  --quiet
+
+echo "✓ Cloud Function deployed: ${FUNCTION_NAME}"
+
+# --- IAM: Cloud Function SA trenger run.admin ---
+PROJECT_NUMBER=$(gcloud projects describe "${PROJECT_ID}" --format="value(projectNumber)")
+gcloud projects add-iam-policy-binding "${PROJECT_ID}" \
+  --member="serviceAccount:${PROJECT_NUMBER}-compute@developer.gserviceaccount.com" \
+  --role="roles/run.admin" --quiet
+gcloud projects add-iam-policy-binding "${PROJECT_ID}" \
+  --member="serviceAccount:${PROJECT_NUMBER}-compute@developer.gserviceaccount.com" \
+  --role="roles/secretmanager.secretAccessor" --quiet
+echo "✓ IAM: run.admin + secretmanager.secretAccessor gitt til Cloud Function SA"
+
+echo ""
+echo "=== 10: Cost Guard COMPLETE ==="
+echo "  Legg til webhook URL i Secret Manager:"
+echo "    gcloud secrets create webhook-url --data-file=<(echo -n \"https://DIN-WEBHOOK-URL\")"
+echo "  Kompatibel med: Google Chat, Slack, Discord, Teams"
+echo "  Tester: publiser en test-melding på Pub/Sub:"
+echo "    gcloud pubsub topics publish ${PUBSUB_TOPIC} --message='{\"budgetAmount\":75,\"costAmount\":75,\"alertThresholdExceeded\":1.0}'"
--- a/infrastructure/notifications/webhook-setup.md
+++ b/infrastructure/notifications/webhook-setup.md
@ -0,0 +1,51 @@
+# Webhook / Chat-app oppsett
+
+Budsjett-alerts og deploy-notifikasjoner sendes til en valgfri chat-app
+via en enkel webhook-URL lagret i Secret Manager.
+
+## Kompatible apper
+
+| App | Webhook-format | Instruksjon |
+|---|---|---|
+| Google Chat | `{"text": "..."}` | Rom → Apps → Webhooks → Legg til webhook |
+| Slack | `{"text": "..."}` | Apps → Incoming Webhooks → Add to Slack |
+| Discord | `{"content": "..."}` | Kanal-innstillinger → Integrations → Webhooks |
+| Teams | `{"text": "..."}` | Kanal → Connectors → Incoming Webhook |
+
+> **Discord-merk:** Discord bruker `content` i stedet for `text`. Oppdater
+> `main.py` i `10-cost-guard.sh` og `cloudbuild.yaml` notify-steget til
+> `{"content": "$MSG"}` hvis du bruker Discord.
+
+## Legge til webhook URL
+
+```bash
+# Ny secret
+gcloud secrets create webhook-url \
+  --data-file=<(echo -n "https://DIN-WEBHOOK-URL") \
+  --project=propane-will-491900-m5
+
+# Oppdater eksisterende secret
+gcloud secrets versions add webhook-url \
+  --data-file=<(echo -n "https://NY-WEBHOOK-URL") \
+  --project=propane-will-491900-m5
+```
+
+## Hva utløser varsler
+
+| Hendelse | Avsender | Terskel |
+|---|---|---|
+| Dev-budsjett 70% nådd | Cloud Function (`10-cost-guard.sh`) | $52.50 av $75 |
+| Dev-budsjett 90% nådd | Cloud Function | $67.50 av $75 |
+| Prod-budsjett 50% nådd | Cloud Function | $250 av $500 |
+| Prod-budsjett 80% nådd | Cloud Function | $400 av $500 |
+| Budsjett 100% — auto-teardown | Cloud Function | Cloud Run stoppes |
+| Deploy fullført | `cloudbuild.yaml` steg 5 | Hver push til main |
+
+## Teste varsling manuelt
+
+```bash
+# Simuler 100% budsjett-alert (trigger auto-teardown)
+gcloud pubsub topics publish billing-alert-auto-teardown \
+  --message='{"budgetAmount":75,"costAmount":75,"alertThresholdExceeded":1.0}' \
+  --project=propane-will-491900-m5
+```