Make EKS e2e less flaky.

Author: s3rj1k

test/e2e/data/e2e_eks_conf.yaml

@@ -149,7 +149,7 @@ intervals:
   default/wait-machine-status: ["20m", "10s"]
   default/wait-infra-subnets: ["5m", "30s"]
   default/wait-control-plane-upgrade: ["35m", "30s"]
-  default/wait-addon-status: ["10m", "30s"]
+  default/wait-addon-status: ["15m", "30s"]
   default/wait-create-identity: ["1m", "10s"]
   default/wait-deployment-ready: ["5m", "10s"]
   default/wait-loadbalancer-ready: ["5m", "30s"]

test/e2e/shared/suite.go

@@ -145,7 +145,7 @@ func Node1BeforeSuite(e2eCtx *E2EContext) []byte {
 				success = false
 			}
 			return success
-		}, 10*time.Minute, 5*time.Second).Should(BeTrue(), "Should've eventually succeeded creating an AWS CloudFormation stack")
+		}, 30*time.Minute, 5*time.Second).Should(BeTrue(), "Should've eventually succeeded creating an AWS CloudFormation stack")
 	}
 
 	ensureStackTags(e2eCtx.AWSSession, bootstrapTemplate.Spec.StackName, bootstrapTags)

test/e2e/suites/managed/eks_test.go

@@ -22,6 +22,7 @@ package managed
 import (
 	"context"
 	"fmt"
+	"time"
 
 	"github.com/onsi/ginkgo/v2"
 	. "github.com/onsi/gomega"
@@ -76,15 +77,22 @@ var _ = ginkgo.Describe("[managed] [general] EKS cluster tests", func() {
 		})
 
 		ginkgo.By("should set environment variables on the aws-node daemonset")
-		CheckAwsNodeEnvVarsSet(ctx, func() UpdateAwsNodeVersionSpecInput {
-			return UpdateAwsNodeVersionSpecInput{
-				E2EConfig:             e2eCtx.E2EConfig,
-				BootstrapClusterProxy: e2eCtx.Environment.BootstrapClusterProxy,
-				AWSSession:            e2eCtx.BootstrapUserAWSSession,
-				Namespace:             namespace,
-				ClusterName:           clusterName,
-			}
-		})
+		Eventually(func() error {
+			defer ginkgo.GinkgoRecover()
+			CheckAwsNodeEnvVarsSet(ctx, func() UpdateAwsNodeVersionSpecInput {
+				return UpdateAwsNodeVersionSpecInput{
+					E2EConfig:             e2eCtx.E2EConfig,
+					BootstrapClusterProxy: e2eCtx.Environment.BootstrapClusterProxy,
+					AWSSession:            e2eCtx.BootstrapUserAWSSession,
+					Namespace:             namespace,
+					ClusterName:           clusterName,
+				}
+			})
+			return nil
+		}).WithTimeout(5*time.Minute).WithPolling(10*time.Second).WithContext(ctx).Should(
+			Succeed(),
+			"Failed to verify AWS Node environment variables after 5 minutes of retries",
+		)
 
 		ginkgo.By("should have the VPC CNI installed")
 		CheckAddonExistsSpec(ctx, func() CheckAddonExistsSpecInput {