Set default api gateway value to none if disabled cluster-wide

cli/local/deploy.go

@@ -62,7 +62,7 @@ func Deploy(env cliconfig.Environment, configPath string, projectFileList []stri
 		}
 	}
 
-	apiConfigs, err := spec.ExtractAPIConfigs(configBytes, types.LocalProviderType, configFileName)
+	apiConfigs, err := spec.ExtractAPIConfigs(configBytes, types.LocalProviderType, configFileName, nil)
 	if err != nil {
 		return schema.DeployResponse{}, err
 	}

docs/deployments/batch-api/api-configuration.md

@@ -20,7 +20,7 @@ Reference the section below which corresponds to your Predictor type: [Python](#
     env: <string: string>  # dictionary of environment variables
   networking:
     endpoint: <string>  # the endpoint for the API (default: <api_name>)
-    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the load balancer will be accessed directly) (default: public)
+    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the API will still be accessible via the load balancer) (default: public, unless disabled cluster-wide)
   compute:
     cpu: <string | int | float>  # CPU request per worker, e.g. 200m or 1 (200m is equivalent to 0.2) (default: 200m)
     gpu: <int>  # GPU request per worker (default: 0)
@@ -55,7 +55,7 @@ See additional documentation for [compute](../compute.md), [networking](../netwo
     env: <string: string>  # dictionary of environment variables
   networking:
     endpoint: <string>  # the endpoint for the API (default: <api_name>)
-    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the load balancer will be accessed directly) (default: public)
+    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the API will still be accessible via the load balancer) (default: public, unless disabled cluster-wide)
   compute:
     cpu: <string | int | float>  # CPU request per worker, e.g. 200m or 1 (200m is equivalent to 0.2) (default: 200m)
     gpu: <int>  # GPU request per worker (default: 0)
@@ -85,7 +85,7 @@ See additional documentation for [compute](../compute.md), [networking](../netwo
     env: <string: string>  # dictionary of environment variables
   networking:
     endpoint: <string>  # the endpoint for the API (default: <api_name>)
-    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the load balancer will be accessed directly) (default: public)
+    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the API will still be accessible via the load balancer) (default: public, unless disabled cluster-wide)
   compute:
     cpu: <string | int | float>  # CPU request per worker, e.g. 200m or 1 (200m is equivalent to 0.2) (default: 200m)
     gpu: <int>  # GPU request per worker (default: 0)

docs/deployments/realtime-api/api-configuration.md

@@ -23,7 +23,7 @@ Reference the section below which corresponds to your Predictor type: [Python](#
   networking:
     endpoint: <string>  # the endpoint for the API (aws only) (default: <api_name>)
     local_port: <int>  # specify the port for API (local only) (default: 8888)
-    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the load balancer will be accessed directly) (default: public)
+    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the API will still be accessible via the load balancer) (default: public, unless disabled cluster-wide)
   compute:
     cpu: <string | int | float>  # CPU request per replica, e.g. 200m or 1 (200m is equivalent to 0.2) (default: 200m)
     gpu: <int>  # GPU request per replica (default: 0)
@@ -80,7 +80,7 @@ See additional documentation for [parallelism](parallelism.md), [autoscaling](au
   networking:
     endpoint: <string>  # the endpoint for the API (aws only) (default: <api_name>)
     local_port: <int>  # specify the port for API (local only) (default: 8888)
-    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the load balancer will be accessed directly) (default: public)
+    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the API will still be accessible via the load balancer) (default: public, unless disabled cluster-wide)
   compute:
     cpu: <string | int | float>  # CPU request per replica, e.g. 200m or 1 (200m is equivalent to 0.2) (default: 200m)
     gpu: <int>  # GPU request per replica (default: 0)
@@ -132,7 +132,7 @@ See additional documentation for [parallelism](parallelism.md), [autoscaling](au
   networking:
     endpoint: <string>  # the endpoint for the API (aws only) (default: <api_name>)
     local_port: <int>  # specify the port for API (local only) (default: 8888)
-    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the load balancer will be accessed directly) (default: public)
+    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the API will still be accessible via the load balancer) (default: public, unless disabled cluster-wide)
   compute:
     cpu: <string | int | float>  # CPU request per replica, e.g. 200m or 1 (200m is equivalent to 0.2) (default: 200m)
     gpu: <int>  # GPU request per replica (default: 0)

docs/deployments/realtime-api/traffic-splitter.md

@@ -17,7 +17,7 @@ Traffic Splitter expects the target Realtime APIs to already be running or be in
   kind: TrafficSplitter  # must be "TrafficSplitter", create an Traffic Splitter which routes traffic to multiple Realtime APIs
   networking:
     endpoint: <string>  # the endpoint for the Traffic Splitter (default: <api_name>)
-    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the load balancer will be accessed directly) (default: public)
+    api_gateway: public | none  # whether to create a public API Gateway endpoint for this API (if not, the API will still be accessible via the load balancer) (default: public, unless disabled cluster-wide)
   apis:  # list of Realtime APIs to target
     - name: <string>  # name of a Realtime API that is already running or is included in the same configuration file (required)
       weight: <int>   # percentage of traffic to route to the Realtime API (all weights must sum to 100) (required)

pkg/operator/resources/errors.go

@@ -95,13 +95,8 @@ func ErrorAPIsNotDeployed(notDeployedAPIs []string) error {
 }
 
 func ErrorAPIGatewayDisabled(apiGatewayType userconfig.APIGatewayType) error {
-	msg := fmt.Sprintf("%s is not permitted because api gateway is disabled cluster-wide", s.UserStr(apiGatewayType))
-	if apiGatewayType == userconfig.PublicAPIGatewayType {
-		msg += fmt.Sprintf(" (%s is the default value, and the valid values are %s)", s.UserStr(userconfig.PublicAPIGatewayType), s.UserStrsAnd(userconfig.APIGatewayTypeStrings()))
-	}
-
 	return errors.WithStack(&errors.Error{
 		Kind:    ErrAPIGatewayDisabled,
-		Message: msg,
+		Message: fmt.Sprintf("%s is not permitted because api gateway is disabled cluster-wide (valid values are %s)", s.UserStr(apiGatewayType), s.UserStrsAnd(userconfig.APIGatewayTypeStrings())),
 	})
 }

pkg/operator/resources/resources.go

@@ -86,7 +86,7 @@ func Deploy(projectBytes []byte, configFileName string, configBytes []byte, forc
 		ConfigFileName: configFileName,
 	}
 
-	apiConfigs, err := spec.ExtractAPIConfigs(configBytes, types.AWSProviderType, configFileName)
+	apiConfigs, err := spec.ExtractAPIConfigs(configBytes, types.AWSProviderType, configFileName, &config.Cluster.Config)
 	if err != nil {
 		return nil, err
 	}

pkg/types/spec/validations.go

@@ -41,19 +41,25 @@ import (
 	libtime "github.com/cortexlabs/cortex/pkg/lib/time"
 	"github.com/cortexlabs/cortex/pkg/lib/urls"
 	"github.com/cortexlabs/cortex/pkg/types"
+	"github.com/cortexlabs/cortex/pkg/types/clusterconfig"
 	"github.com/cortexlabs/cortex/pkg/types/userconfig"
 	kresource "k8s.io/apimachinery/pkg/api/resource"
 )
 
 var AutoscalingTickInterval = 10 * time.Second
 
-func apiValidation(provider types.ProviderType, resource userconfig.Resource) *cr.StructValidation {
+func apiValidation(
+	provider types.ProviderType,
+	resource userconfig.Resource,
+	clusterConfig *clusterconfig.Config, // should be omitted if running locally
+) *cr.StructValidation {
+
 	structFieldValidations := []*cr.StructFieldValidation{}
 	switch resource.Kind {
 	case userconfig.RealtimeAPIKind:
 		structFieldValidations = append(resourceStructValidations,
 			predictorValidation(),
-			networkingValidation(resource.Kind),
+			networkingValidation(resource.Kind, clusterConfig),
 			computeValidation(provider),
 			monitoringValidation(),
 			autoscalingValidation(provider),
@@ -62,13 +68,13 @@ func apiValidation(provider types.ProviderType, resource userconfig.Resource) *c
 	case userconfig.BatchAPIKind:
 		structFieldValidations = append(resourceStructValidations,
 			predictorValidation(),
-			networkingValidation(resource.Kind),
+			networkingValidation(resource.Kind, clusterConfig),
 			computeValidation(provider),
 		)
 	case userconfig.TrafficSplitterKind:
 		structFieldValidations = append(resourceStructValidations,
 			multiAPIsValidation(),
-			networkingValidation(resource.Kind),
+			networkingValidation(resource.Kind, clusterConfig),
 		)
 	}
 	return &cr.StructValidation{
@@ -253,7 +259,16 @@ func monitoringValidation() *cr.StructFieldValidation {
 	}
 }
 
-func networkingValidation(kind userconfig.Kind) *cr.StructFieldValidation {
+func networkingValidation(
+	kind userconfig.Kind,
+	clusterConfig *clusterconfig.Config, // should be omitted if running locally
+) *cr.StructFieldValidation {
+
+	defaultAPIGatewayType := userconfig.PublicAPIGatewayType
+	if clusterConfig != nil && clusterConfig.APIGatewaySetting == clusterconfig.NoneAPIGatewaySetting {
+		defaultAPIGatewayType = userconfig.NoneAPIGatewayType
+	}
+
 	structFieldValidation := []*cr.StructFieldValidation{
 		{
 			StructField: "Endpoint",
@@ -266,7 +281,7 @@ func networkingValidation(kind userconfig.Kind) *cr.StructFieldValidation {
 			StructField: "APIGateway",
 			StringValidation: &cr.StringValidation{
 				AllowedValues: userconfig.APIGatewayTypeStrings(),
-				Default:       userconfig.PublicAPIGatewayType.String(),
+				Default:       defaultAPIGatewayType.String(),
 			},
 			Parser: func(str string) (interface{}, error) {
 				return userconfig.APIGatewayTypeFromString(str), nil
@@ -570,7 +585,13 @@ var resourceStructValidation = cr.StructValidation{
 	StructFieldValidations: resourceStructValidations,
 }
 
-func ExtractAPIConfigs(configBytes []byte, provider types.ProviderType, configFileName string) ([]userconfig.API, error) {
+func ExtractAPIConfigs(
+	configBytes []byte,
+	provider types.ProviderType,
+	configFileName string,
+	clusterConfig *clusterconfig.Config, // should be omitted if running locally
+) ([]userconfig.API, error) {
+
 	var err error
 
 	configData, err := cr.ReadYAMLBytes(configBytes)
@@ -607,7 +628,7 @@ func ExtractAPIConfigs(configBytes []byte, provider types.ProviderType, configFi
 			}
 		}
 
-		errs = cr.Struct(&api, data, apiValidation(provider, resourceStruct))
+		errs = cr.Struct(&api, data, apiValidation(provider, resourceStruct, clusterConfig))
 		if errors.HasError(errs) {
 			name, _ := data[userconfig.NameKey].(string)
 			kindString, _ := data[userconfig.KindKey].(string)