712 lines
18 KiB
Go
712 lines
18 KiB
Go
package google
|
|
|
|
import (
|
|
"fmt"
|
|
"log"
|
|
"net"
|
|
"regexp"
|
|
|
|
"github.com/hashicorp/terraform/helper/resource"
|
|
"github.com/hashicorp/terraform/helper/schema"
|
|
"google.golang.org/api/container/v1"
|
|
)
|
|
|
|
var (
|
|
instanceGroupManagerURL = regexp.MustCompile("^https://www.googleapis.com/compute/v1/projects/([a-z][a-z0-9-]{5}(?:[-a-z0-9]{0,23}[a-z0-9])?)/zones/([a-z0-9-]*)/instanceGroupManagers/([^/]*)")
|
|
)
|
|
|
|
func resourceContainerCluster() *schema.Resource {
|
|
return &schema.Resource{
|
|
Create: resourceContainerClusterCreate,
|
|
Read: resourceContainerClusterRead,
|
|
Update: resourceContainerClusterUpdate,
|
|
Delete: resourceContainerClusterDelete,
|
|
|
|
Schema: map[string]*schema.Schema{
|
|
"master_auth": &schema.Schema{
|
|
Type: schema.TypeList,
|
|
Optional: true,
|
|
ForceNew: true,
|
|
MaxItems: 1,
|
|
Computed: true,
|
|
Elem: &schema.Resource{
|
|
Schema: map[string]*schema.Schema{
|
|
"client_certificate": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Computed: true,
|
|
},
|
|
"client_key": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Computed: true,
|
|
Sensitive: true,
|
|
},
|
|
"cluster_ca_certificate": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Computed: true,
|
|
},
|
|
"password": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Required: true,
|
|
ForceNew: true,
|
|
Sensitive: true,
|
|
},
|
|
"username": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Required: true,
|
|
ForceNew: true,
|
|
},
|
|
},
|
|
},
|
|
},
|
|
|
|
"name": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Required: true,
|
|
ForceNew: true,
|
|
ValidateFunc: func(v interface{}, k string) (ws []string, errors []error) {
|
|
value := v.(string)
|
|
|
|
if len(value) > 40 {
|
|
errors = append(errors, fmt.Errorf(
|
|
"%q cannot be longer than 40 characters", k))
|
|
}
|
|
if !regexp.MustCompile("^[a-z0-9-]+$").MatchString(value) {
|
|
errors = append(errors, fmt.Errorf(
|
|
"%q can only contain lowercase letters, numbers and hyphens", k))
|
|
}
|
|
if !regexp.MustCompile("^[a-z]").MatchString(value) {
|
|
errors = append(errors, fmt.Errorf(
|
|
"%q must start with a letter", k))
|
|
}
|
|
if !regexp.MustCompile("[a-z0-9]$").MatchString(value) {
|
|
errors = append(errors, fmt.Errorf(
|
|
"%q must end with a number or a letter", k))
|
|
}
|
|
return
|
|
},
|
|
},
|
|
|
|
"zone": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Required: true,
|
|
ForceNew: true,
|
|
},
|
|
|
|
"initial_node_count": &schema.Schema{
|
|
Type: schema.TypeInt,
|
|
Optional: true,
|
|
ForceNew: true,
|
|
},
|
|
|
|
"additional_zones": &schema.Schema{
|
|
Type: schema.TypeList,
|
|
Optional: true,
|
|
Computed: true,
|
|
ForceNew: true,
|
|
Elem: &schema.Schema{Type: schema.TypeString},
|
|
},
|
|
|
|
"cluster_ipv4_cidr": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Optional: true,
|
|
Computed: true,
|
|
ForceNew: true,
|
|
ValidateFunc: func(v interface{}, k string) (ws []string, errors []error) {
|
|
value := v.(string)
|
|
_, ipnet, err := net.ParseCIDR(value)
|
|
|
|
if err != nil || ipnet == nil || value != ipnet.String() {
|
|
errors = append(errors, fmt.Errorf(
|
|
"%q must contain a valid CIDR", k))
|
|
}
|
|
return
|
|
},
|
|
},
|
|
|
|
"description": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Optional: true,
|
|
ForceNew: true,
|
|
},
|
|
|
|
"endpoint": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Computed: true,
|
|
},
|
|
|
|
"instance_group_urls": &schema.Schema{
|
|
Type: schema.TypeList,
|
|
Computed: true,
|
|
Elem: &schema.Schema{Type: schema.TypeString},
|
|
},
|
|
|
|
"logging_service": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Optional: true,
|
|
Computed: true,
|
|
ForceNew: true,
|
|
},
|
|
|
|
"monitoring_service": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Optional: true,
|
|
Computed: true,
|
|
ForceNew: true,
|
|
},
|
|
|
|
"network": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Optional: true,
|
|
Default: "default",
|
|
ForceNew: true,
|
|
},
|
|
"subnetwork": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Optional: true,
|
|
ForceNew: true,
|
|
},
|
|
"addons_config": &schema.Schema{
|
|
Type: schema.TypeList,
|
|
Optional: true,
|
|
ForceNew: true,
|
|
MaxItems: 1,
|
|
Elem: &schema.Resource{
|
|
Schema: map[string]*schema.Schema{
|
|
"http_load_balancing": &schema.Schema{
|
|
Type: schema.TypeList,
|
|
Optional: true,
|
|
ForceNew: true,
|
|
MaxItems: 1,
|
|
Elem: &schema.Resource{
|
|
Schema: map[string]*schema.Schema{
|
|
"disabled": &schema.Schema{
|
|
Type: schema.TypeBool,
|
|
Optional: true,
|
|
ForceNew: true,
|
|
},
|
|
},
|
|
},
|
|
},
|
|
"horizontal_pod_autoscaling": &schema.Schema{
|
|
Type: schema.TypeList,
|
|
Optional: true,
|
|
ForceNew: true,
|
|
MaxItems: 1,
|
|
Elem: &schema.Resource{
|
|
Schema: map[string]*schema.Schema{
|
|
"disabled": &schema.Schema{
|
|
Type: schema.TypeBool,
|
|
Optional: true,
|
|
ForceNew: true,
|
|
},
|
|
},
|
|
},
|
|
},
|
|
},
|
|
},
|
|
},
|
|
"node_config": &schema.Schema{
|
|
Type: schema.TypeList,
|
|
Optional: true,
|
|
Computed: true,
|
|
ForceNew: true,
|
|
Elem: &schema.Resource{
|
|
Schema: map[string]*schema.Schema{
|
|
"machine_type": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Optional: true,
|
|
Computed: true,
|
|
ForceNew: true,
|
|
},
|
|
|
|
"disk_size_gb": &schema.Schema{
|
|
Type: schema.TypeInt,
|
|
Optional: true,
|
|
Computed: true,
|
|
ForceNew: true,
|
|
ValidateFunc: func(v interface{}, k string) (ws []string, errors []error) {
|
|
value := v.(int)
|
|
|
|
if value < 10 {
|
|
errors = append(errors, fmt.Errorf(
|
|
"%q cannot be less than 10", k))
|
|
}
|
|
return
|
|
},
|
|
},
|
|
|
|
"local_ssd_count": &schema.Schema{
|
|
Type: schema.TypeInt,
|
|
Optional: true,
|
|
Computed: true,
|
|
ForceNew: true,
|
|
ValidateFunc: func(v interface{}, k string) (ws []string, errors []error) {
|
|
value := v.(int)
|
|
|
|
if value < 0 {
|
|
errors = append(errors, fmt.Errorf(
|
|
"%q cannot be negative", k))
|
|
}
|
|
return
|
|
},
|
|
},
|
|
|
|
"oauth_scopes": &schema.Schema{
|
|
Type: schema.TypeList,
|
|
Optional: true,
|
|
Computed: true,
|
|
ForceNew: true,
|
|
Elem: &schema.Schema{
|
|
Type: schema.TypeString,
|
|
StateFunc: func(v interface{}) string {
|
|
return canonicalizeServiceScope(v.(string))
|
|
},
|
|
},
|
|
},
|
|
|
|
"service_account": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Optional: true,
|
|
Computed: true,
|
|
ForceNew: true,
|
|
},
|
|
|
|
"metadata": &schema.Schema{
|
|
Type: schema.TypeMap,
|
|
Optional: true,
|
|
ForceNew: true,
|
|
Elem: schema.TypeString,
|
|
},
|
|
|
|
"image_type": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Optional: true,
|
|
Computed: true,
|
|
ForceNew: true,
|
|
},
|
|
},
|
|
},
|
|
},
|
|
|
|
"node_version": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Optional: true,
|
|
Computed: true,
|
|
},
|
|
|
|
"node_pool": &schema.Schema{
|
|
Type: schema.TypeList,
|
|
Optional: true,
|
|
Computed: true,
|
|
ForceNew: true, // TODO(danawillow): Add ability to add/remove nodePools
|
|
Elem: &schema.Resource{
|
|
Schema: map[string]*schema.Schema{
|
|
"initial_node_count": &schema.Schema{
|
|
Type: schema.TypeInt,
|
|
Required: true,
|
|
ForceNew: true,
|
|
},
|
|
|
|
"name": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Optional: true,
|
|
Computed: true,
|
|
ConflictsWith: []string{"node_pool.name_prefix"},
|
|
ForceNew: true,
|
|
},
|
|
|
|
"name_prefix": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Optional: true,
|
|
ForceNew: true,
|
|
},
|
|
},
|
|
},
|
|
},
|
|
|
|
"project": &schema.Schema{
|
|
Type: schema.TypeString,
|
|
Optional: true,
|
|
ForceNew: true,
|
|
},
|
|
},
|
|
}
|
|
}
|
|
|
|
func resourceContainerClusterCreate(d *schema.ResourceData, meta interface{}) error {
|
|
config := meta.(*Config)
|
|
|
|
project, err := getProject(d, config)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
zoneName := d.Get("zone").(string)
|
|
clusterName := d.Get("name").(string)
|
|
|
|
cluster := &container.Cluster{
|
|
Name: clusterName,
|
|
InitialNodeCount: int64(d.Get("initial_node_count").(int)),
|
|
}
|
|
|
|
if v, ok := d.GetOk("master_auth"); ok {
|
|
masterAuths := v.([]interface{})
|
|
masterAuth := masterAuths[0].(map[string]interface{})
|
|
cluster.MasterAuth = &container.MasterAuth{
|
|
Password: masterAuth["password"].(string),
|
|
Username: masterAuth["username"].(string),
|
|
}
|
|
}
|
|
|
|
if v, ok := d.GetOk("node_version"); ok {
|
|
cluster.InitialClusterVersion = v.(string)
|
|
}
|
|
|
|
if v, ok := d.GetOk("additional_zones"); ok {
|
|
locationsList := v.([]interface{})
|
|
locations := []string{}
|
|
for _, v := range locationsList {
|
|
location := v.(string)
|
|
locations = append(locations, location)
|
|
if location == zoneName {
|
|
return fmt.Errorf("additional_zones should not contain the original 'zone'.")
|
|
}
|
|
}
|
|
locations = append(locations, zoneName)
|
|
cluster.Locations = locations
|
|
}
|
|
|
|
if v, ok := d.GetOk("cluster_ipv4_cidr"); ok {
|
|
cluster.ClusterIpv4Cidr = v.(string)
|
|
}
|
|
|
|
if v, ok := d.GetOk("description"); ok {
|
|
cluster.Description = v.(string)
|
|
}
|
|
|
|
if v, ok := d.GetOk("logging_service"); ok {
|
|
cluster.LoggingService = v.(string)
|
|
}
|
|
|
|
if v, ok := d.GetOk("monitoring_service"); ok {
|
|
cluster.MonitoringService = v.(string)
|
|
}
|
|
|
|
if _, ok := d.GetOk("network"); ok {
|
|
network, err := getNetworkName(d, "network")
|
|
if err != nil {
|
|
return err
|
|
}
|
|
cluster.Network = network
|
|
}
|
|
|
|
if v, ok := d.GetOk("subnetwork"); ok {
|
|
cluster.Subnetwork = v.(string)
|
|
}
|
|
|
|
if v, ok := d.GetOk("addons_config"); ok {
|
|
addonsConfig := v.([]interface{})[0].(map[string]interface{})
|
|
cluster.AddonsConfig = &container.AddonsConfig{}
|
|
|
|
if v, ok := addonsConfig["http_load_balancing"]; ok && len(v.([]interface{})) > 0 {
|
|
addon := v.([]interface{})[0].(map[string]interface{})
|
|
cluster.AddonsConfig.HttpLoadBalancing = &container.HttpLoadBalancing{
|
|
Disabled: addon["disabled"].(bool),
|
|
}
|
|
}
|
|
|
|
if v, ok := addonsConfig["horizontal_pod_autoscaling"]; ok && len(v.([]interface{})) > 0 {
|
|
addon := v.([]interface{})[0].(map[string]interface{})
|
|
cluster.AddonsConfig.HorizontalPodAutoscaling = &container.HorizontalPodAutoscaling{
|
|
Disabled: addon["disabled"].(bool),
|
|
}
|
|
}
|
|
}
|
|
if v, ok := d.GetOk("node_config"); ok {
|
|
nodeConfigs := v.([]interface{})
|
|
if len(nodeConfigs) > 1 {
|
|
return fmt.Errorf("Cannot specify more than one node_config.")
|
|
}
|
|
nodeConfig := nodeConfigs[0].(map[string]interface{})
|
|
|
|
cluster.NodeConfig = &container.NodeConfig{}
|
|
|
|
if v, ok = nodeConfig["machine_type"]; ok {
|
|
cluster.NodeConfig.MachineType = v.(string)
|
|
}
|
|
|
|
if v, ok = nodeConfig["disk_size_gb"]; ok {
|
|
cluster.NodeConfig.DiskSizeGb = int64(v.(int))
|
|
}
|
|
|
|
if v, ok = nodeConfig["local_ssd_count"]; ok {
|
|
cluster.NodeConfig.LocalSsdCount = int64(v.(int))
|
|
}
|
|
|
|
if v, ok := nodeConfig["oauth_scopes"]; ok {
|
|
scopesList := v.([]interface{})
|
|
scopes := []string{}
|
|
for _, v := range scopesList {
|
|
scopes = append(scopes, canonicalizeServiceScope(v.(string)))
|
|
}
|
|
|
|
cluster.NodeConfig.OauthScopes = scopes
|
|
}
|
|
|
|
if v, ok = nodeConfig["service_account"]; ok {
|
|
cluster.NodeConfig.ServiceAccount = v.(string)
|
|
}
|
|
|
|
if v, ok = nodeConfig["metadata"]; ok {
|
|
m := make(map[string]string)
|
|
for k, val := range v.(map[string]interface{}) {
|
|
m[k] = val.(string)
|
|
}
|
|
cluster.NodeConfig.Metadata = m
|
|
}
|
|
|
|
if v, ok = nodeConfig["image_type"]; ok {
|
|
cluster.NodeConfig.ImageType = v.(string)
|
|
}
|
|
}
|
|
|
|
nodePoolsCount := d.Get("node_pool.#").(int)
|
|
if nodePoolsCount > 0 {
|
|
nodePools := make([]*container.NodePool, 0, nodePoolsCount)
|
|
for i := 0; i < nodePoolsCount; i++ {
|
|
prefix := fmt.Sprintf("node_pool.%d", i)
|
|
|
|
nodeCount := d.Get(prefix + ".initial_node_count").(int)
|
|
|
|
var name string
|
|
if v, ok := d.GetOk(prefix + ".name"); ok {
|
|
name = v.(string)
|
|
} else if v, ok := d.GetOk(prefix + ".name_prefix"); ok {
|
|
name = resource.PrefixedUniqueId(v.(string))
|
|
} else {
|
|
name = resource.UniqueId()
|
|
}
|
|
|
|
nodePool := &container.NodePool{
|
|
Name: name,
|
|
InitialNodeCount: int64(nodeCount),
|
|
}
|
|
|
|
nodePools = append(nodePools, nodePool)
|
|
}
|
|
cluster.NodePools = nodePools
|
|
}
|
|
|
|
req := &container.CreateClusterRequest{
|
|
Cluster: cluster,
|
|
}
|
|
|
|
op, err := config.clientContainer.Projects.Zones.Clusters.Create(
|
|
project, zoneName, req).Do()
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
// Wait until it's created
|
|
waitErr := containerOperationWait(config, op, project, zoneName, "creating GKE cluster", 30, 3)
|
|
if waitErr != nil {
|
|
// The resource didn't actually create
|
|
d.SetId("")
|
|
return waitErr
|
|
}
|
|
|
|
log.Printf("[INFO] GKE cluster %s has been created", clusterName)
|
|
|
|
d.SetId(clusterName)
|
|
|
|
return resourceContainerClusterRead(d, meta)
|
|
}
|
|
|
|
func resourceContainerClusterRead(d *schema.ResourceData, meta interface{}) error {
|
|
config := meta.(*Config)
|
|
|
|
project, err := getProject(d, config)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
zoneName := d.Get("zone").(string)
|
|
|
|
cluster, err := config.clientContainer.Projects.Zones.Clusters.Get(
|
|
project, zoneName, d.Get("name").(string)).Do()
|
|
if err != nil {
|
|
return handleNotFoundError(err, d, fmt.Sprintf("Container Cluster %q", d.Get("name").(string)))
|
|
}
|
|
|
|
d.Set("name", cluster.Name)
|
|
d.Set("zone", cluster.Zone)
|
|
|
|
locations := []string{}
|
|
if len(cluster.Locations) > 1 {
|
|
for _, location := range cluster.Locations {
|
|
if location != cluster.Zone {
|
|
locations = append(locations, location)
|
|
}
|
|
}
|
|
}
|
|
d.Set("additional_zones", locations)
|
|
|
|
d.Set("endpoint", cluster.Endpoint)
|
|
|
|
masterAuth := []map[string]interface{}{
|
|
map[string]interface{}{
|
|
"username": cluster.MasterAuth.Username,
|
|
"password": cluster.MasterAuth.Password,
|
|
"client_certificate": cluster.MasterAuth.ClientCertificate,
|
|
"client_key": cluster.MasterAuth.ClientKey,
|
|
"cluster_ca_certificate": cluster.MasterAuth.ClusterCaCertificate,
|
|
},
|
|
}
|
|
d.Set("master_auth", masterAuth)
|
|
|
|
d.Set("initial_node_count", cluster.InitialNodeCount)
|
|
d.Set("node_version", cluster.CurrentNodeVersion)
|
|
d.Set("cluster_ipv4_cidr", cluster.ClusterIpv4Cidr)
|
|
d.Set("description", cluster.Description)
|
|
d.Set("logging_service", cluster.LoggingService)
|
|
d.Set("monitoring_service", cluster.MonitoringService)
|
|
d.Set("network", d.Get("network").(string))
|
|
d.Set("subnetwork", cluster.Subnetwork)
|
|
d.Set("node_config", flattenClusterNodeConfig(cluster.NodeConfig))
|
|
d.Set("node_pool", flattenClusterNodePools(d, cluster.NodePools))
|
|
|
|
if igUrls, err := getInstanceGroupUrlsFromManagerUrls(config, cluster.InstanceGroupUrls); err != nil {
|
|
return err
|
|
} else {
|
|
d.Set("instance_group_urls", igUrls)
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
func resourceContainerClusterUpdate(d *schema.ResourceData, meta interface{}) error {
|
|
config := meta.(*Config)
|
|
|
|
project, err := getProject(d, config)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
zoneName := d.Get("zone").(string)
|
|
clusterName := d.Get("name").(string)
|
|
desiredNodeVersion := d.Get("node_version").(string)
|
|
|
|
req := &container.UpdateClusterRequest{
|
|
Update: &container.ClusterUpdate{
|
|
DesiredNodeVersion: desiredNodeVersion,
|
|
},
|
|
}
|
|
op, err := config.clientContainer.Projects.Zones.Clusters.Update(
|
|
project, zoneName, clusterName, req).Do()
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
// Wait until it's updated
|
|
waitErr := containerOperationWait(config, op, project, zoneName, "updating GKE cluster", 10, 2)
|
|
if waitErr != nil {
|
|
return waitErr
|
|
}
|
|
|
|
log.Printf("[INFO] GKE cluster %s has been updated to %s", d.Id(),
|
|
desiredNodeVersion)
|
|
|
|
return resourceContainerClusterRead(d, meta)
|
|
}
|
|
|
|
func resourceContainerClusterDelete(d *schema.ResourceData, meta interface{}) error {
|
|
config := meta.(*Config)
|
|
|
|
project, err := getProject(d, config)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
zoneName := d.Get("zone").(string)
|
|
clusterName := d.Get("name").(string)
|
|
|
|
log.Printf("[DEBUG] Deleting GKE cluster %s", d.Get("name").(string))
|
|
op, err := config.clientContainer.Projects.Zones.Clusters.Delete(
|
|
project, zoneName, clusterName).Do()
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
// Wait until it's deleted
|
|
waitErr := containerOperationWait(config, op, project, zoneName, "deleting GKE cluster", 10, 3)
|
|
if waitErr != nil {
|
|
return waitErr
|
|
}
|
|
|
|
log.Printf("[INFO] GKE cluster %s has been deleted", d.Id())
|
|
|
|
d.SetId("")
|
|
|
|
return nil
|
|
}
|
|
|
|
// container engine's API currently mistakenly returns the instance group manager's
|
|
// URL instead of the instance group's URL in its responses. This shim detects that
|
|
// error, and corrects it, by fetching the instance group manager URL and retrieving
|
|
// the instance group manager, then using that to look up the instance group URL, which
|
|
// is then substituted.
|
|
//
|
|
// This should be removed when the API response is fixed.
|
|
func getInstanceGroupUrlsFromManagerUrls(config *Config, igmUrls []string) ([]string, error) {
|
|
instanceGroupURLs := make([]string, 0, len(igmUrls))
|
|
for _, u := range igmUrls {
|
|
if !instanceGroupManagerURL.MatchString(u) {
|
|
instanceGroupURLs = append(instanceGroupURLs, u)
|
|
continue
|
|
}
|
|
matches := instanceGroupManagerURL.FindStringSubmatch(u)
|
|
instanceGroupManager, err := config.clientCompute.InstanceGroupManagers.Get(matches[1], matches[2], matches[3]).Do()
|
|
if err != nil {
|
|
return nil, fmt.Errorf("Error reading instance group manager returned as an instance group URL: %s", err)
|
|
}
|
|
instanceGroupURLs = append(instanceGroupURLs, instanceGroupManager.InstanceGroup)
|
|
}
|
|
return instanceGroupURLs, nil
|
|
}
|
|
|
|
func flattenClusterNodeConfig(c *container.NodeConfig) []map[string]interface{} {
|
|
config := []map[string]interface{}{
|
|
map[string]interface{}{
|
|
"machine_type": c.MachineType,
|
|
"disk_size_gb": c.DiskSizeGb,
|
|
"local_ssd_count": c.LocalSsdCount,
|
|
"service_account": c.ServiceAccount,
|
|
"metadata": c.Metadata,
|
|
"image_type": c.ImageType,
|
|
},
|
|
}
|
|
|
|
if len(c.OauthScopes) > 0 {
|
|
config[0]["oauth_scopes"] = c.OauthScopes
|
|
}
|
|
|
|
return config
|
|
}
|
|
|
|
func flattenClusterNodePools(d *schema.ResourceData, c []*container.NodePool) []map[string]interface{} {
|
|
count := len(c)
|
|
|
|
nodePools := make([]map[string]interface{}, 0, count)
|
|
|
|
for i, np := range c {
|
|
nodePool := map[string]interface{}{
|
|
"name": np.Name,
|
|
"name_prefix": d.Get(fmt.Sprintf("node_pool.%d.name_prefix", i)),
|
|
"initial_node_count": np.InitialNodeCount,
|
|
}
|
|
nodePools = append(nodePools, nodePool)
|
|
}
|
|
|
|
return nodePools
|
|
}
|