[1.15] Add --wait to helm tests and other improvements to support CD tools

changelog/v1.15.9/helm-wait.yaml

@@ -0,0 +1,4 @@
+changelog:
+  - type: NON_USER_FACING
+    description: Most CI tools wait until resources are ready to mark a release as successful such as ArgoCD and Flux. We can emulate this by passing the --wait flag to helm rather than individually adding support for every CD tool out there.
+

install/helm/gloo/templates/0-namespace.yaml

@@ -14,4 +14,5 @@ metadata:
     {{- end}}
   annotations:
     "helm.sh/hook": pre-install
+    "helm.sh/hook-weight": "-1" # Ensure that this is always the first hook triggered
 {{- end}}

install/helm/gloo/templates/19-gloo-mtls-certgen-job.yaml

@@ -13,6 +13,8 @@ metadata:
   name: gloo-mtls-certgen
   namespace: {{ .Release.Namespace }}
   annotations:
+    {{- /* With ArgoCD, there is no concept or installs or upgrades. Everything is a sync. */}}
+    {{- /* Due to this behavour of ArgoCD, the value of `gateway.certGenJob.runOnUpdate` can not be respected and will run on every sync. */}}
     {{- if .Values.gateway.certGenJob.runOnUpdate }}
     "helm.sh/hook": pre-install, pre-upgrade
     {{- include "gloo.jobHelmDeletePolicySucceededAndBeforeCreation" .Values.gateway.certGenJob | nindent 4 }}

test/kube2e/helm/helm_test.go

@@ -529,6 +529,17 @@ func installGloo(testHelper *helper.SoloTestHelper, chartUri string, fromRelease
 		args = append(args, chartUri)
 	}
 	args = append(args, "-n", testHelper.InstallNamespace,
+		// As most CD tools wait for resources to be ready before marking the release as successful,
+		// we're emulating that here by passing these two flags.
+		// This way we ensure that we indirectly add support for CD tools
+		"--wait",
+		"--wait-for-jobs",
+		// We run our e2e tests on a kind cluster, but kind hasn’t implemented LoadBalancer support.
+		// This leads to the service being in a pending state.
+		// Since the --wait flag is set, this can cause the upgrade to fail
+		// as helm waits until the service is ready and eventually times out.
+		// So instead we use the service type as ClusterIP to work around this limitation.
+		"--set", "gatewayProxies.gatewayProxy.service.type=ClusterIP",
 		"--create-namespace",
 		"--values", helmValuesFile)
 	if strictValidation {
@@ -566,6 +577,17 @@ func upgradeGloo(testHelper *helper.SoloTestHelper, chartUri string, crdDir stri
 	defer cleanupFunc()
 
 	var args = []string{"upgrade", testHelper.HelmChartName,
+		// As most CD tools wait for resources to be ready before marking the release as successful,
+		// we're emulating that here by passing these two flags.
+		// This way we ensure that we indirectly add support for CD tools
+		"--wait",
+		"--wait-for-jobs",
+		// We run our e2e tests on a kind cluster, but kind hasn’t implemented LoadBalancer support.
+		// This leads to the service being in a pending state.
+		// Since the --wait flag is set, this can cause the upgrade to fail
+		// as helm waits until the service is ready and eventually times out.
+		// So instead we use the service type as ClusterIP to work around this limitation.
+		"--set", "gatewayProxies.gatewayProxy.service.type=ClusterIP",
 		"-n", testHelper.InstallNamespace,
 		"--values", valueOverrideFile}
 	if targetRelease != "" {

test/kube2e/upgrade/upgrade_test.go

@@ -275,6 +275,17 @@ func installGloo(testHelper *helper.SoloTestHelper, fromRelease string, strictVa
 		"--version", fromRelease)
 
 	args = append(args, "-n", testHelper.InstallNamespace,
+		// As most CD tools wait for resources to be ready before marking the release as successful,
+		// we're emulating that here by passing these two flags.
+		// This way we ensure that we indirectly add support for CD tools
+		"--wait",
+		"--wait-for-jobs",
+		// We run our e2e tests on a kind cluster, but kind hasn’t implemented LoadBalancer support.
+		// This leads to the service being in a pending state.
+		// Since the --wait flag is set, this can cause the upgrade to fail
+		// as helm waits until the service is ready and eventually times out.
+		// So instead we use the service type as ClusterIP to work around this limitation.
+		"--set", "gatewayProxies.gatewayProxy.service.type=ClusterIP",
 		"--create-namespace",
 		"--values", helmValuesFile)
 	if strictValidation {
@@ -306,6 +317,17 @@ func upgradeGloo(testHelper *helper.SoloTestHelper, chartUri string, targetRelea
 	defer cleanupFunc()
 
 	var args = []string{"upgrade", testHelper.HelmChartName, chartUri,
+		// As most CD tools wait for resources to be ready before marking the release as successful,
+		// we're emulating that here by passing these two flags.
+		// This way we ensure that we indirectly add support for CD tools
+		"--wait",
+		"--wait-for-jobs",
+		// We run our e2e tests on a kind cluster, but kind hasn’t implemented LoadBalancer support.
+		// This leads to the service being in a pending state.
+		// Since the --wait flag is set, this can cause the upgrade to fail
+		// as helm waits until the service is ready and eventually times out.
+		// So instead we use the service type as ClusterIP to work around this limitation.
+		"--set", "gatewayProxies.gatewayProxy.service.type=ClusterIP",
 		"-n", testHelper.InstallNamespace,
 		"--values", valueOverrideFile}
 	if targetReleasedVersion != "" {