feat: Support for tenant with no base cluster topology (#199)

This is the companion of
kurtosis-tech/kardinal-kontrol#16

.github/workflows/ci-e2e-tests.yml

@@ -178,3 +178,26 @@ jobs:
       - name: Delete template
         run: |
           KARDINAL_CLI_DEV_MODE=TRUE /tmp/kardinal-cli template delete extra-item-shared
+
+      - name: Delete base topology and dev flows
+        run: |
+          KARDINAL_CLI_DEV_MODE=TRUE /tmp/kardinal-cli flow delete prod > kardinal.out
+          cat kardinal.out
+          if KARDINAL_CLI_DEV_MODE=TRUE /tmp/kardinal-cli flow ls | grep prod; then echo "Topologies not deleted"; exit 1; fi
+          
+          # Check that the services have been terminated
+          while [ "$(kubectl get pods -n prod 2>&1 >/dev/null)" != "No resources found in prod namespace." ]
+          do
+              echo "Waiting for the services to terminate..."
+              kubectl get pods -n prod
+              ((c++)) && ((c==12)) && exit 1
+              sleep 10
+          done
+          kubectl get pods -n prod
+          
+          # Check for errors in the kardinal manager logs
+          if kubectl logs -n default -l dev.kardinal.app-id=kardinal-manager | grep "ERRO"
+          then
+            echo "Errors found in the kardinal manager"
+            kubectl logs -n default -l dev.kardinal.app-id=kardinal-manager | grep "ERRO"
+          fi

kardinal-manager/kardinal-manager/cluster_manager/cluster_manager.go

@@ -284,6 +284,12 @@ func (manager *ClusterManager) CleanUpClusterResources(ctx context.Context, clus
 	servicesByNS := lo.GroupBy(*clusterResources.Services, func(item corev1.Service) string {
 		return item.Namespace
 	})
+	if len(servicesByNS) == 0 {
+		// There are no resources to apply so we attempt to clear resources in the namespace set in the dummy gateway
+		// sent by the kontrol service.  This happens when the tenant has no base cluster topology; no initial deploy
+		// or the topologies have been deleted.
+		servicesByNS[clusterResources.Gateway.GetNamespace()] = []corev1.Service{}
+	}
 	for namespace, services := range servicesByNS {
 		if err := manager.cleanUpServicesInNamespace(ctx, namespace, services); err != nil {
 			return stacktrace.Propagate(err, "An error occurred cleaning up services '%+v' in namespace '%s'", services, namespace)
@@ -292,6 +298,9 @@ func (manager *ClusterManager) CleanUpClusterResources(ctx context.Context, clus
 
 	// Clean up deployments
 	deploymentsByNS := lo.GroupBy(*clusterResources.Deployments, func(item appsv1.Deployment) string { return item.Namespace })
+	if len(deploymentsByNS) == 0 {
+		deploymentsByNS[clusterResources.Gateway.GetNamespace()] = []appsv1.Deployment{}
+	}
 	for namespace, deployments := range deploymentsByNS {
 		if err := manager.cleanUpDeploymentsInNamespace(ctx, namespace, deployments); err != nil {
 			return stacktrace.Propagate(err, "An error occurred cleaning up deployments '%+v' in namespace '%s'", deployments, namespace)
@@ -300,6 +309,9 @@ func (manager *ClusterManager) CleanUpClusterResources(ctx context.Context, clus
 
 	// Clean up virtual services
 	virtualServicesByNS := lo.GroupBy(*clusterResources.VirtualServices, func(item v1alpha3.VirtualService) string { return item.Namespace })
+	if len(virtualServicesByNS) == 0 {
+		virtualServicesByNS[clusterResources.Gateway.GetNamespace()] = []v1alpha3.VirtualService{}
+	}
 	for namespace, virtualServices := range virtualServicesByNS {
 		if err := manager.cleanUpVirtualServicesInNamespace(ctx, namespace, virtualServices); err != nil {
 			return stacktrace.Propagate(err, "An error occurred cleaning up virtual services '%+v' in namespace '%s'", virtualServices, namespace)
@@ -310,6 +322,9 @@ func (manager *ClusterManager) CleanUpClusterResources(ctx context.Context, clus
 	destinationRulesByNS := lo.GroupBy(*clusterResources.DestinationRules, func(item v1alpha3.DestinationRule) string {
 		return item.Namespace
 	})
+	if len(destinationRulesByNS) == 0 {
+		destinationRulesByNS[clusterResources.Gateway.GetNamespace()] = []v1alpha3.DestinationRule{}
+	}
 	for namespace, destinationRules := range destinationRulesByNS {
 		if err := manager.cleanUpDestinationRulesInNamespace(ctx, namespace, destinationRules); err != nil {
 			return stacktrace.Propagate(err, "An error occurred cleaning up destination rules '%+v' in namespace '%s'", destinationRules, namespace)
@@ -331,6 +346,9 @@ func (manager *ClusterManager) CleanUpClusterResources(ctx context.Context, clus
 		envoyFiltersByNS := lo.GroupBy(*clusterResources.EnvoyFilters, func(item v1alpha3.EnvoyFilter) string {
 			return item.Namespace
 		})
+		if len(envoyFiltersByNS) == 0 {
+			envoyFiltersByNS[clusterResources.Gateway.GetNamespace()] = []v1alpha3.EnvoyFilter{}
+		}
 		for namespace, envoyFilters := range envoyFiltersByNS {
 			if err := manager.cleanupEnvoyFiltersInNamespace(ctx, namespace, envoyFilters); err != nil {
 				return stacktrace.Propagate(err, "An error occurred cleaning up envoy filters '%+v' in namespace '%s'", envoyFilters, namespace)
@@ -343,6 +361,9 @@ func (manager *ClusterManager) CleanUpClusterResources(ctx context.Context, clus
 		authorizationPoliciesByNS := lo.GroupBy(*clusterResources.AuthorizationPolicies, func(item securityv1beta1.AuthorizationPolicy) string {
 			return item.Namespace
 		})
+		if len(authorizationPoliciesByNS) == 0 {
+			authorizationPoliciesByNS[clusterResources.Gateway.GetNamespace()] = []securityv1beta1.AuthorizationPolicy{}
+		}
 		for namespace, authorizationPolicies := range authorizationPoliciesByNS {
 			if err := manager.cleanupAuthorizationPoliciesInNamespace(ctx, namespace, authorizationPolicies); err != nil {
 				return stacktrace.Propagate(err, "An error occurred cleaning up authorization policies '%+v' in namespace '%s'", authorizationPolicies, namespace)

kardinal-manager/kardinal-manager/main.go

@@ -13,7 +13,6 @@ import (
 const (
 	successExitCode                = 0
 	clusterConfigEndpointEnvVarKey = "KARDINAL_MANAGER_CLUSTER_CONFIG_ENDPOINT"
-	tenantUuidEnvVarKey            = "KARDINAL_MANAGER_TENANT_UUID"
 )
 
 func main() {