gcp.vertex.workbench.idle

Author: javvaji-devops

CHANGELOG.md

@@ -57,18 +57,18 @@ cleancloud scan --provider aws --category all
      - utilisation_pct: 1.2
      - estimated_monthly_cost: ~$23 374/mois
 
-2. [GCP] Instance Workbench GPU inactive (>14 jours, 31 jours sans activité)
-   Risque     : Critique
+2. [GCP] Endpoint Vertex AI inactif (2 réplique(s) toujours active(s), zéro requête)
+   Risque     : Élevé
    Confiance  : High
-   Ressource  : gcp.vertex.workbench.instance → projects/ml-platform/locations/us-central1/instances/research-nb-gpu
+   Ressource  : gcp.vertex.endpoint → projects/ml-platform/locations/us-central1/endpoints/8842531067721654272
    Région     : us-central1
-   Règle      : gcp.vertex.workbench.idle
-   Raison     : Instance Workbench sans activité sur le plan de contrôle depuis 31 jours, état ACTIVE
+   Règle      : gcp.vertex.endpoint.idle
+   Raison     : L'endpoint a un plancher de service provisionné de 2 réplique(s) ; la télémétrie de comptage des requêtes (couverture : complète) montre un taux maximum observé == 0 sur une fenêtre de 14j
    Détails :
-     - machine_type: a2-highgpu-4g
-     - accelerator_type: NVIDIA_TESLA_A100
-     - accelerator_count: 4
-     - estimated_monthly_cost: ~$11 732/mois
+     - provisioned_serving_floor: 2
+     - in_scope_model_count: 1
+     - has_accelerator: true
+     - telemetry_coverage_state: complete
 
 3. [Azure] Instance de calcul Azure ML inactive (31 jours sans activité)
    Risque     : Élevé

README.fr.md

@@ -57,18 +57,18 @@ Found 10 findings (6 hygiene + 4 AI/ML):
      - utilisation_pct: 1.2
      - estimated_monthly_cost: ~$23,374/month
 
-2. [GCP] Idle GPU-Backed Workbench Instance (>14 Days Idle, 31 Days Since Activity)
-   Risk       : Critical
+2. [GCP] Idle Vertex AI Endpoint (2 replica(s) always on, zero requests)
+   Risk       : High
    Confidence : High
-   Resource   : gcp.vertex.workbench.instance → projects/ml-platform/locations/us-central1/instances/research-nb-gpu
+   Resource   : gcp.vertex.endpoint → projects/ml-platform/locations/us-central1/endpoints/8842531067721654272
    Region     : us-central1
-   Rule       : gcp.vertex.workbench.idle
-   Reason     : Workbench instance has had no control-plane activity for 31 days while ACTIVE
+   Rule       : gcp.vertex.endpoint.idle
+   Reason     : Endpoint has provisioned serving floor of 2 replica(s); request-count telemetry (coverage: complete) shows max observed rate == 0 over 14d window
    Details:
-     - machine_type: a2-highgpu-4g
-     - accelerator_type: NVIDIA_TESLA_A100
-     - accelerator_count: 4
-     - estimated_monthly_cost: ~$11,732/month
+     - provisioned_serving_floor: 2
+     - in_scope_model_count: 1
+     - has_accelerator: true
+     - telemetry_coverage_state: complete
 
 3. [Azure] Idle Azure ML Compute Instance (31 Days Since Last Activity)
    Risk       : High