Merge pull request #893 from NVIDIA/nse-err

NVIDIA · Aug 2, 2024 · eb7e633 · eb7e633
2 parents 2993021 + 199da11
commit eb7e633
Show file tree

Hide file tree

Showing 3 changed files with 13 additions and 0 deletions.
diff --git a/assets/state-node-status-exporter/0200_role.yaml b/assets/state-node-status-exporter/0200_role.yaml
@@ -19,3 +19,11 @@ rules:
   verbs:
   - get
   - list
+- apiGroups:
+  - apps
+  resources:
+  - daemonsets
+  verbs:
+  - get
+  - list
+  - watch
diff --git a/assets/state-node-status-exporter/0700_daemonset.yaml b/assets/state-node-status-exporter/0700_daemonset.yaml
@@ -40,6 +40,10 @@ spec:
           valueFrom:
             fieldRef:
               fieldPath: spec.nodeName
+        - name: OPERATOR_NAMESPACE
+          valueFrom:
+            fieldRef:
+              fieldPath: metadata.namespace
         ports:
         - name: node-status
           containerPort: 8000

diff --git a/validator/metrics.go b/validator/metrics.go
@@ -238,6 +238,7 @@ func (nm *NodeMetrics) watchDriverValidation() {
 			nm.driverValidation.Set(1)
 			nm.driverValidationLastSuccess.Set(float64(time.Now().Unix()))
 		} else {
+			log.Errorf("failed to validate driver: %v", err)
 			nm.driverValidation.Set(0)
 		}
 		time.Sleep(driverValidationCheckDelaySeconds * time.Second)