Yuyu3/remove relay

src/k8s-extension/HISTORY.rst

@@ -18,7 +18,7 @@ Release History
 1.1.0
 ++++++++++++++++++
 * Migrate Extensions api-version to 2022-03-01
-* microsoft.azureml.kubernetes: Remove inference private review warning message
+* microsoft.azureml.kubernetes: Remove inference private review warning message, disable service bus by default
 * microsoft.openservicemesh: Enable System-assigned identity
 
 1.0.4

src/k8s-extension/azext_k8s_extension/custom.py

@@ -297,6 +297,7 @@ def update_k8s_extension(
         version,
         config_settings,
         config_protected_settings,
+        extension,
         yes,
     )
 

src/k8s-extension/azext_k8s_extension/partner_extensions/AzureMLKubernetes.py

@@ -57,6 +57,7 @@ def __init__(self):
         self.AZURE_LOG_ANALYTICS_CONNECTION_STRING = 'azure_log_analytics.connection_string'
         self.JOB_SCHEDULER_LOCATION_KEY = 'jobSchedulerLocation'
         self.CLUSTER_NAME_FRIENDLY_KEY = 'cluster_name_friendly'
+        self.NGINX_INGRESS_ENABLED_KEY = 'nginxIngress.enabled'
 
         # component flag
         self.ENABLE_TRAINING = 'enableTraining'
@@ -66,6 +67,10 @@ def __init__(self):
         self.RELAY_SERVER_CONNECTION_STRING = 'relayServerConnectionString'  # create relay connection string if None
         self.SERVICE_BUS_CONNECTION_STRING = 'serviceBusConnectionString'  # create service bus if None
         self.LOG_ANALYTICS_WS_ENABLED = 'logAnalyticsWS'  # create log analytics workspace if true
+        # default to false when creating the extension
+        self.SERVICE_BUS_ENABLED = 'servicebus.enabled'
+        # default to false if cluster is AKS when creating the extension
+        self.RELAY_SERVER_ENABLED = 'relayserver.enabled'
 
         # constants for azure resources creation
         self.RELAY_HC_AUTH_NAME = 'azureml_rw'
@@ -107,8 +112,8 @@ def Create(self, cmd, client, resource_group_name, cluster_name, name, cluster_t
         if scope == 'namespace':
             raise InvalidArgumentValueError("Invalid scope '{}'.  This extension can't be installed "
                                             "only at 'cluster' scope.".format(scope))
-        if not release_namespace:
-            release_namespace = self.DEFAULT_RELEASE_NAMESPACE
+        # set release name explicitly to azureml
+        release_namespace = self.DEFAULT_RELEASE_NAMESPACE
         scope_cluster = ScopeCluster(release_namespace=release_namespace)
         ext_scope = Scope(cluster=scope_cluster, namespace=None)
 
@@ -136,6 +141,8 @@ def Create(self, cmd, client, resource_group_name, cluster_name, name, cluster_t
                         nodeCount += agent['count']
                     if nodeCount < 3:
                         configuration_settings['clusterPurpose'] = 'DevTest'
+                if resource.properties.get('distribution', '').lower() == self.OPEN_SHIFT:
+                    configuration_settings[self.OPEN_SHIFT] = 'true'
             except:
                 pass
         except CloudError as ex:
@@ -150,6 +157,20 @@ def Create(self, cmd, client, resource_group_name, cluster_name, name, cluster_t
             self.JOB_SCHEDULER_LOCATION_KEY, cluster_location)
         configuration_settings[self.CLUSTER_NAME_FRIENDLY_KEY] = configuration_settings.get(
             self.CLUSTER_NAME_FRIENDLY_KEY, cluster_name)
+        # do not enable service bus by default
+        configuration_settings[self.SERVICE_BUS_ENABLED] = configuration_settings.get(self.SERVICE_BUS_ENABLED, 'false')
+
+        # do not enable relay for managed cluster(AKS) by default, do not enable nginx for ARC by default
+        if cluster_type == "managedClusters":
+            configuration_settings[self.RELAY_SERVER_ENABLED] = configuration_settings.get(self.RELAY_SERVER_ENABLED,
+                                                                                           'false')
+            configuration_settings[self.NGINX_INGRESS_ENABLED_KEY] = configuration_settings.get(
+                self.NGINX_INGRESS_ENABLED_KEY, 'true')
+        else:
+            configuration_settings[self.RELAY_SERVER_ENABLED] = configuration_settings.get(self.RELAY_SERVER_ENABLED,
+                                                                                           'true')
+            configuration_settings[self.NGINX_INGRESS_ENABLED_KEY] = configuration_settings.get(
+                self.NGINX_INGRESS_ENABLED_KEY, 'false')
 
         # create Azure resources need by the extension based on the config.
         self.__create_required_resource(
@@ -186,7 +207,7 @@ def Delete(self, cmd, client, resource_group_name, cluster_name, name, cluster_t
         user_confirmation_factory(cmd, yes)
 
     def Update(self, cmd, resource_group_name, cluster_name, auto_upgrade_minor_version, release_train, version, configuration_settings,
-               configuration_protected_settings, yes=False):
+               configuration_protected_settings, original_extension, yes=False):
         self.__normalize_config(configuration_settings, configuration_protected_settings)
 
         # Prompt message to ask customer to confirm again
@@ -281,7 +302,8 @@ def Update(self, cmd, resource_group_name, cluster_name, auto_upgrade_minor_vers
                 except azure.core.exceptions.HttpResponseError:
                     logger.info("Failed to get log analytics connection string.")
 
-            if self.RELAY_SERVER_CONNECTION_STRING not in configuration_protected_settings:
+            if original_extension.configuration_settings.get(self.RELAY_SERVER_ENABLED).lower() != 'false' \
+                    and self.RELAY_SERVER_CONNECTION_STRING not in configuration_protected_settings:
                 try:
                     relay_connection_string, _, _ = _get_relay_connection_str(
                         cmd, subscription_id, resource_group_name, cluster_name, '', self.RELAY_HC_AUTH_NAME, True)
@@ -292,7 +314,8 @@ def Update(self, cmd, resource_group_name, cluster_name, auto_upgrade_minor_vers
                         raise ResourceNotFoundError("Relay server not found.") from ex
                     raise AzureResponseError("Failed to get relay connection string.") from ex
 
-            if self.SERVICE_BUS_CONNECTION_STRING not in configuration_protected_settings:
+            if original_extension.configuration_settings.get(self.SERVICE_BUS_ENABLED).lower() != 'false' \
+                    and self.SERVICE_BUS_CONNECTION_STRING not in configuration_protected_settings:
                 try:
                     service_bus_connection_string, _ = _get_service_bus_connection_string(
                         cmd, subscription_id, resource_group_name, cluster_name, '', {}, True)
@@ -494,7 +517,8 @@ def __create_required_resource(
             configuration_settings[self.AZURE_LOG_ANALYTICS_CUSTOMER_ID_KEY] = ws_costumer_id
             configuration_protected_settings[self.AZURE_LOG_ANALYTICS_CONNECTION_STRING] = shared_key
 
-        if not configuration_settings.get(self.RELAY_SERVER_CONNECTION_STRING) and \
+        if str(configuration_settings.get(self.RELAY_SERVER_ENABLED)).lower() != 'false' and \
+                not configuration_settings.get(self.RELAY_SERVER_CONNECTION_STRING) and \
                 not configuration_protected_settings.get(self.RELAY_SERVER_CONNECTION_STRING):
             logger.info('==== BEGIN RELAY CREATION ====')
             relay_connection_string, hc_resource_id, hc_name = _get_relay_connection_str(
@@ -504,7 +528,8 @@ def __create_required_resource(
             configuration_settings[self.HC_RESOURCE_ID_KEY] = hc_resource_id
             configuration_settings[self.RELAY_HC_NAME_KEY] = hc_name
 
-        if not configuration_settings.get(self.SERVICE_BUS_CONNECTION_STRING) and \
+        if str(configuration_settings.get(self.SERVICE_BUS_ENABLED)).lower() != 'false' and \
+                not configuration_settings.get(self.SERVICE_BUS_CONNECTION_STRING) and \
                 not configuration_protected_settings.get(self.SERVICE_BUS_CONNECTION_STRING):
             logger.info('==== BEGIN SERVICE BUS CREATION ====')
             topic_sub_mapping = {

src/k8s-extension/azext_k8s_extension/partner_extensions/DefaultExtension.py

@@ -71,6 +71,7 @@ def Update(
         version,
         configuration_settings,
         configuration_protected_settings,
+        original_extension: Extension,
         yes=False,
     ):
         """Default validations & defaults for Update

src/k8s-extension/azext_k8s_extension/partner_extensions/PartnerExtensionModel.py

@@ -43,6 +43,7 @@ def Update(
         version: str,
         configuration_settings: dict,
         configuration_protected_settings: dict,
+        original_extension: Extension,
         yes: bool,
     ) -> PatchExtension:
         pass

testing/test/extensions/public/AzureMLKubernetes.Tests.ps1

@@ -58,11 +58,8 @@ Describe 'AzureML Kubernetes Testing' {
     It "Deletes the extension from the cluster with inference enabled" {
         # cleanup the relay and servicebus
         $relayResourceID = Get-ExtensionConfigurationSettings $extensionName $relayResourceIDKey
-        $serviceBusResourceID = Get-ExtensionConfigurationSettings $extensionName $serviceBusResourceIDKey
         $relayNamespaceName = $relayResourceID.split("/")[8]
-        $serviceBusNamespaceName = $serviceBusResourceID.split("/")[8]
         az relay namespace delete --resource-group $ENVCONFIG.resourceGroup --name $relayNamespaceName
-        az servicebus namespace delete --resource-group $ENVCONFIG.resourceGroup --name $serviceBusNamespaceName
 
         $output = az $Env:K8sExtensionName delete -c $($ENVCONFIG.arcClusterName) -g $($ENVCONFIG.resourceGroup) --cluster-type connectedClusters -n $extensionName --force
         $? | Should -BeTrue