Add readiness probes (#312)

trustyai-explainability · Oct 14, 2024 · 7979767 · 7979767
1 parent 91237c5
commit 7979767
Show file tree

Hide file tree

Showing 3 changed files with 20 additions and 2 deletions.
diff --git a/controllers/database.go b/controllers/database.go
@@ -12,7 +12,6 @@ import (
 	"sigs.k8s.io/controller-runtime/pkg/client"
 )
 
-// checkDatabaseAccessible checks if the TrustyAI service pod failed with database issues.
 func (r *TrustyAIServiceReconciler) checkDatabaseAccessible(ctx context.Context, instance *trustyaiopendatahubiov1alpha1.TrustyAIService) (bool, error) {
 	deployment := &appsv1.Deployment{}
 	err := r.Get(ctx, types.NamespacedName{Name: instance.Name, Namespace: instance.Namespace}, deployment)

diff --git a/controllers/statuses.go b/controllers/statuses.go
@@ -87,7 +87,12 @@ func (r *TrustyAIServiceReconciler) reconcileStatuses(ctx context.Context, insta
 			UpdateRouteAvailable(saved)
 
 			if instance.Spec.Storage.IsStorageDatabase() || instance.IsMigration() {
-				UpdateDBAvailable(saved)
+				if status.DBReady {
+					UpdateDBAvailable(saved)
+				} else {
+					UpdateDBConnectionError(saved)
+					return
+				}
 			}
 
 			UpdateTrustyAIServiceAvailable(saved)

diff --git a/controllers/templates/service/deployment.tmpl.yaml b/controllers/templates/service/deployment.tmpl.yaml
@@ -122,6 +122,20 @@ spec:
             - name: STORAGE_MIGRATION_CONFIG_FROM_FILENAME
               value: {{ .Instance.Spec.Data.Filename }}
             {{ end }}
+          readinessProbe:
+            httpGet:
+              path: /q/health/ready
+              port: 8080
+            initialDelaySeconds: 10
+            periodSeconds: 5
+            timeoutSeconds: 2
+          livenessProbe:
+            httpGet:
+              path: /q/health/live
+              port: 8080
+            initialDelaySeconds: 10
+            periodSeconds: 5
+            timeoutSeconds: 2
           volumeMounts:
             - name: {{ .Instance.Name }}-internal
               readOnly: false