Add osSku field to NodePoolConfig (#209)

Author: johnstairs

cli/internal/install/cloudinstall/cloudconfig.go

@@ -98,6 +98,7 @@ type ClusterConfig struct {
 type NodePoolConfig struct {
 	Name     string `json:"name"`
 	VMSize   string `json:"vmSize"`
+	OsSku    string `json:"osSku"`
 	MinCount int32  `json:"minCount"`
 	MaxCount int32  `json:"maxCount"`
 }

cli/internal/install/cloudinstall/compute.go

@@ -143,7 +143,7 @@ func (inst *Installer) createCluster(ctx context.Context, clusterConfig *Cluster
 			MinCount:            &clusterConfig.SystemNodePool.MinCount,
 			MaxCount:            &clusterConfig.SystemNodePool.MaxCount,
 			OSType:              Ptr(armcontainerservice.OSTypeLinux),
-			OSSKU:               Ptr(armcontainerservice.OSSKUAzureLinux),
+			OSSKU:               Ptr(armcontainerservice.OSSKU(clusterConfig.SystemNodePool.OsSku)),
 			Tags:                tags,
 		},
 	}
@@ -159,7 +159,7 @@ func (inst *Installer) createCluster(ctx context.Context, clusterConfig *Cluster
 			MinCount:            &np.MinCount,
 			MaxCount:            &np.MaxCount,
 			OSType:              Ptr(armcontainerservice.OSTypeLinux),
-			OSSKU:               Ptr(armcontainerservice.OSSKUAzureLinux),
+			OSSKU:               Ptr(armcontainerservice.OSSKU(np.OsSku)),
 			NodeLabels: map[string]*string{
 				"tyger": Ptr("run"),
 			},
@@ -293,6 +293,11 @@ func (inst *Installer) createCluster(ctx context.Context, clusterConfig *Cluster
 
 		poller, err = clustersClient.BeginCreateOrUpdate(ctx, inst.Config.Cloud.ResourceGroup, clusterConfig.Name, cluster, nil)
 		if err != nil {
+			var respErr *azcore.ResponseError
+			if errors.As(err, &respErr) && respErr.ErrorCode == "InvalidOSSKU" {
+				return nil, fmt.Errorf("failed to create cluster: adjust the `osSku` field on the node pool: %w", err)
+			}
+
 			return nil, fmt.Errorf("failed to create cluster: %w", err)
 		}
 

cli/internal/install/cloudinstall/config.tpl

@@ -30,10 +30,12 @@ cloud:
             vmSize: Standard_DS12_v2
             minCount: {{ .CpuNodePoolMinCount }}
             maxCount: 10
+            # osSku: defaults to AzureLinux
           - name: gpunp
             vmSize: Standard_NC6s_v3
             minCount: {{ .GpuNodePoolMinCount }}
             maxCount: 10
+            # osSku: defaults to AzureLinux
 
     # These are the principals that will have the ability to run `tyger api install`.
     # They will have access to the "tyger" namespace in each cluster and will have

cli/internal/install/cloudinstall/validation.go

@@ -200,6 +200,15 @@ func quickValidateNodePoolConfig(success *bool, np *NodePoolConfig, minNodeCount
 	if np.MinCount > np.MaxCount {
 		validationError(success, "The `minCount` field must be less than or equal to the `maxCount` field")
 	}
+
+	if np.OsSku == "" {
+		np.OsSku = string(armcontainerservice.OSSKUAzureLinux)
+	} else {
+		supportedValues := []string{string(armcontainerservice.OSSKUAzureLinux), string(armcontainerservice.OSSKUUbuntu)}
+		if !slices.Contains(supportedValues, np.OsSku) {
+			validationError(success, "The `osSku` field must be one of [%s]", strings.Join(supportedValues, ", "))
+		}
+	}
 }
 
 func quickValidateStorageConfig(success *bool, cloudConfig *CloudConfig) {

docs/introduction/installation/cloud-installation.md

@@ -79,10 +79,12 @@ cloud:
             vmSize: Standard_DS12_v2
             minCount: 1
             maxCount: 10
+            # osSku: defaults to AzureLinux
           - name: gpunp
             vmSize: Standard_NC6s_v3
             minCount: 0
             maxCount: 10
+            # osSku: defaults to AzureLinux
 
     # These are the principals that will have the ability to run `tyger api install`.
     # They will have access to the "tyger" namespace in each cluster and will have