fix: handle webhook inability to connect error v34 (#1632)

pkg/operator/operator.go

@@ -237,6 +237,11 @@ func (o *Operator) Start(ctx context.Context, cp cloudprovider.CloudProvider) {
 			ctx = injection.WithClient(ctx, o.GetClient())
 			webhooks.Start(ctx, o.GetConfig(), o.webhooks...)
 		}()
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			webhooks.ValidateConversionEnabled(ctx, o.GetClient())
+		}()
 	}
 	wg.Wait()
 }

pkg/webhooks/webhooks.go

@@ -23,6 +23,7 @@ import (
 	"io"
 	"net/http"
 	"strings"
+	"time"
 
 	"github.com/samber/lo"
 	"go.uber.org/zap"
@@ -41,6 +42,7 @@ import (
 	"knative.dev/pkg/webhook/resourcesemantics"
 	"knative.dev/pkg/webhook/resourcesemantics/conversion"
 	"knative.dev/pkg/webhook/resourcesemantics/validation"
+	"sigs.k8s.io/controller-runtime/pkg/client"
 	"sigs.k8s.io/controller-runtime/pkg/healthz"
 
 	v1 "sigs.k8s.io/karpenter/pkg/apis/v1"
@@ -206,3 +208,22 @@ func HealthProbe(ctx context.Context) healthz.Checker {
 		return nil
 	}
 }
+
+func ValidateConversionEnabled(ctx context.Context, kubeclient client.Client) {
+	// allow context to exist longer than cache sync timeout which has a default of 120 seconds
+	listCtx, cancel := context.WithTimeout(ctx, 130*time.Second)
+	defer cancel()
+	var err error
+	v1np := &v1.NodePoolList{}
+	for {
+		err = kubeclient.List(listCtx, v1np, &client.ListOptions{Limit: 1})
+		if err == nil {
+			return
+		}
+		select {
+		case <-listCtx.Done():
+			panic("Conversion webhook enabled but unable to complete call: " + err.Error())
+		case <-time.After(10 * time.Second):
+		}
+	}
+}