published on Wednesday, Nov 18, 2026 by Pulumi
published on Wednesday, Nov 18, 2026 by Pulumi
A Vertex AI ModelDevelopmentCluster resource that manages a cluster of compute resources with an orchestrator (such as Slurm or GKE) for model development and training workloads.
Warning: This resource is in beta, and should be used with the terraform-provider-google-beta provider. See Provider Versions for more details on beta resources.
To get more information about ModelDevelopmentCluster, see:
- API documentation
- How-to Guides
Example Usage
Vertex Ai Model Development Cluster Basic
import * as pulumi from "@pulumi/pulumi";
import * as gcp from "@pulumi/gcp";
const project = gcp.organizations.getProject({});
const net = new gcp.compute.Network("net", {
name: "mdc-net",
autoCreateSubnetworks: false,
});
const subnet = new gcp.compute.Subnetwork("subnet", {
name: "mdc-subnet",
ipCidrRange: "10.0.0.0/24",
region: "us-east4",
network: net.id,
privateIpGoogleAccess: true,
});
const internal = new gcp.compute.Firewall("internal", {
name: "mdc-fw",
network: net.name,
direction: "INGRESS",
sourceRanges: ["10.0.0.0/24"],
allows: [
{
protocol: "tcp",
ports: ["0-65535"],
},
{
protocol: "udp",
ports: ["0-65535"],
},
{
protocol: "icmp",
},
],
});
const home = new gcp.filestore.Instance("home", {
name: "mdc-home",
location: "us-east4-b",
tier: "BASIC_HDD",
fileShares: {
capacityGb: 1024,
name: "share1",
},
networks: [{
network: net.name,
modes: ["MODE_IPV4"],
}],
});
const data = new gcp.filestore.Instance("data", {
name: "mdc-data",
location: "us-east4-b",
tier: "BASIC_HDD",
fileShares: {
capacityGb: 1024,
name: "share1",
},
networks: [{
network: net.name,
modes: ["MODE_IPV4"],
}],
});
const res = new gcp.compute.Reservation("res", {
name: "mdc-res",
zone: "us-east4-b",
specificReservationRequired: true,
specificReservation: {
count: 1,
instanceProperties: {
machineType: "n2-standard-4",
},
},
});
const _default = new gcp.vertex.AiModelDevelopmentCluster("default", {
region: "us-east4",
modelDevelopmentClusterId: "tftest",
displayName: "tftest",
deletionPolicy: "DELETE",
updateMode: "USER_AND_SERVICE",
labels: {
env: "dev",
},
runtimeSpec: {
serviceAccount: project.then(project => `${project.number}-compute@developer.gserviceaccount.com`),
},
lustreExportConfig: {
destinationBucket: "mdc-backup-bucket",
daysOfWeeks: [
"MONDAY",
"THURSDAY",
],
startTime: {
hours: 4,
minutes: 30,
},
},
network: {
network: net.id,
subnetwork: subnet.id,
},
orchestratorSpec: {
slurmSpec: {
homeDirectoryStorage: home.id,
loginNodePoolId: "login",
defaultPartitionId: "main",
prologBashScripts: [`#!/bin/bash
echo prolog`],
epilogBashScripts: [`#!/bin/bash
echo epilog`],
config: {
PriorityType: "priority/multifactor",
PriorityWeightAge: "1000",
},
nodeSets: [{
nodePoolId: "compute",
config: {
Weight: "10",
},
}],
partitions: [{
id: "main",
nodePoolIds: [
"compute",
"flexpool",
"respool",
],
config: {
PriorityTier: "10",
},
}],
},
},
nodePools: [
{
id: "compute",
zone: "us-east4-b",
provisioningModel: "ON_DEMAND",
nodeImage: "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113",
startupScript: `#!/bin/bash
echo hello`,
enablePublicIps: true,
filestores: [data.id],
labels: {
pool: "compute",
},
bootDisk: {
bootDiskSizeGb: 100,
bootDiskType: "pd-ssd",
},
machineSpec: {
machineType: "n2-standard-4",
},
scalingSpec: {
minNodeCount: 1,
maxNodeCount: 2,
},
},
{
id: "flexpool",
zone: "us-east4-b",
provisioningModel: "FLEX_START",
flexStartMaxDuration: "3600s",
nodeImage: "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113",
startupScript: `#!/bin/bash
echo hello`,
enablePublicIps: true,
filestores: [data.id],
labels: {
pool: "flexpool",
},
bootDisk: {
bootDiskSizeGb: 100,
bootDiskType: "hyperdisk-balanced",
},
machineSpec: {
machineType: "a3-ultragpu-8g",
},
scalingSpec: {
minNodeCount: 0,
maxNodeCount: 1,
},
},
{
id: "login",
zone: "us-east4-b",
provisioningModel: "ON_DEMAND",
nodeImage: "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113",
startupScript: `#!/bin/bash
echo hello`,
enablePublicIps: true,
filestores: [data.id],
labels: {
pool: "login",
},
bootDisk: {
bootDiskSizeGb: 100,
bootDiskType: "pd-ssd",
},
machineSpec: {
machineType: "n2-standard-4",
},
scalingSpec: {
minNodeCount: 1,
maxNodeCount: 1,
},
},
{
id: "respool",
zone: "us-east4-b",
provisioningModel: "RESERVATION",
nodeImage: "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113",
startupScript: `#!/bin/bash
echo hello`,
enablePublicIps: true,
filestores: [data.id],
labels: {
pool: "respool",
},
bootDisk: {
bootDiskSizeGb: 100,
bootDiskType: "pd-ssd",
},
machineSpec: {
machineType: "n2-standard-4",
reservationAffinity: {
reservationAffinityType: "SPECIFIC_RESERVATION",
values: [res.id],
},
},
scalingSpec: {
minNodeCount: 0,
maxNodeCount: 1,
},
},
],
}, {
dependsOn: [internal],
});
import pulumi
import pulumi_gcp as gcp
project = gcp.organizations.get_project()
net = gcp.compute.Network("net",
name="mdc-net",
auto_create_subnetworks=False)
subnet = gcp.compute.Subnetwork("subnet",
name="mdc-subnet",
ip_cidr_range="10.0.0.0/24",
region="us-east4",
network=net.id,
private_ip_google_access=True)
internal = gcp.compute.Firewall("internal",
name="mdc-fw",
network=net.name,
direction="INGRESS",
source_ranges=["10.0.0.0/24"],
allows=[
{
"protocol": "tcp",
"ports": ["0-65535"],
},
{
"protocol": "udp",
"ports": ["0-65535"],
},
{
"protocol": "icmp",
},
])
home = gcp.filestore.Instance("home",
name="mdc-home",
location="us-east4-b",
tier="BASIC_HDD",
file_shares={
"capacity_gb": 1024,
"name": "share1",
},
networks=[{
"network": net.name,
"modes": ["MODE_IPV4"],
}])
data = gcp.filestore.Instance("data",
name="mdc-data",
location="us-east4-b",
tier="BASIC_HDD",
file_shares={
"capacity_gb": 1024,
"name": "share1",
},
networks=[{
"network": net.name,
"modes": ["MODE_IPV4"],
}])
res = gcp.compute.Reservation("res",
name="mdc-res",
zone="us-east4-b",
specific_reservation_required=True,
specific_reservation={
"count": 1,
"instance_properties": {
"machine_type": "n2-standard-4",
},
})
default = gcp.vertex.AiModelDevelopmentCluster("default",
region="us-east4",
model_development_cluster_id="tftest",
display_name="tftest",
deletion_policy="DELETE",
update_mode="USER_AND_SERVICE",
labels={
"env": "dev",
},
runtime_spec={
"service_account": f"{project.number}-compute@developer.gserviceaccount.com",
},
lustre_export_config={
"destination_bucket": "mdc-backup-bucket",
"days_of_weeks": [
"MONDAY",
"THURSDAY",
],
"start_time": {
"hours": 4,
"minutes": 30,
},
},
network={
"network": net.id,
"subnetwork": subnet.id,
},
orchestrator_spec={
"slurm_spec": {
"home_directory_storage": home.id,
"login_node_pool_id": "login",
"default_partition_id": "main",
"prolog_bash_scripts": ["""#!/bin/bash
echo prolog"""],
"epilog_bash_scripts": ["""#!/bin/bash
echo epilog"""],
"config": {
"PriorityType": "priority/multifactor",
"PriorityWeightAge": "1000",
},
"node_sets": [{
"node_pool_id": "compute",
"config": {
"Weight": "10",
},
}],
"partitions": [{
"id": "main",
"node_pool_ids": [
"compute",
"flexpool",
"respool",
],
"config": {
"PriorityTier": "10",
},
}],
},
},
node_pools=[
{
"id": "compute",
"zone": "us-east4-b",
"provisioning_model": "ON_DEMAND",
"node_image": "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113",
"startup_script": """#!/bin/bash
echo hello""",
"enable_public_ips": True,
"filestores": [data.id],
"labels": {
"pool": "compute",
},
"boot_disk": {
"boot_disk_size_gb": 100,
"boot_disk_type": "pd-ssd",
},
"machine_spec": {
"machine_type": "n2-standard-4",
},
"scaling_spec": {
"min_node_count": 1,
"max_node_count": 2,
},
},
{
"id": "flexpool",
"zone": "us-east4-b",
"provisioning_model": "FLEX_START",
"flex_start_max_duration": "3600s",
"node_image": "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113",
"startup_script": """#!/bin/bash
echo hello""",
"enable_public_ips": True,
"filestores": [data.id],
"labels": {
"pool": "flexpool",
},
"boot_disk": {
"boot_disk_size_gb": 100,
"boot_disk_type": "hyperdisk-balanced",
},
"machine_spec": {
"machine_type": "a3-ultragpu-8g",
},
"scaling_spec": {
"min_node_count": 0,
"max_node_count": 1,
},
},
{
"id": "login",
"zone": "us-east4-b",
"provisioning_model": "ON_DEMAND",
"node_image": "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113",
"startup_script": """#!/bin/bash
echo hello""",
"enable_public_ips": True,
"filestores": [data.id],
"labels": {
"pool": "login",
},
"boot_disk": {
"boot_disk_size_gb": 100,
"boot_disk_type": "pd-ssd",
},
"machine_spec": {
"machine_type": "n2-standard-4",
},
"scaling_spec": {
"min_node_count": 1,
"max_node_count": 1,
},
},
{
"id": "respool",
"zone": "us-east4-b",
"provisioning_model": "RESERVATION",
"node_image": "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113",
"startup_script": """#!/bin/bash
echo hello""",
"enable_public_ips": True,
"filestores": [data.id],
"labels": {
"pool": "respool",
},
"boot_disk": {
"boot_disk_size_gb": 100,
"boot_disk_type": "pd-ssd",
},
"machine_spec": {
"machine_type": "n2-standard-4",
"reservation_affinity": {
"reservation_affinity_type": "SPECIFIC_RESERVATION",
"values": [res.id],
},
},
"scaling_spec": {
"min_node_count": 0,
"max_node_count": 1,
},
},
],
opts = pulumi.ResourceOptions(depends_on=[internal]))
package main
import (
"github.com/pulumi/pulumi-gcp/sdk/v10/go/gcp/compute"
"github.com/pulumi/pulumi-gcp/sdk/v10/go/gcp/filestore"
"github.com/pulumi/pulumi-gcp/sdk/v10/go/gcp/organizations"
"github.com/pulumi/pulumi-gcp/sdk/v10/go/gcp/vertex"
"github.com/pulumi/pulumi/sdk/v3/go/pulumi"
)
func main() {
pulumi.Run(func(ctx *pulumi.Context) error {
project, err := organizations.LookupProject(ctx, &organizations.LookupProjectArgs{}, nil)
if err != nil {
return err
}
net, err := compute.NewNetwork(ctx, "net", &compute.NetworkArgs{
Name: pulumi.String("mdc-net"),
AutoCreateSubnetworks: pulumi.Bool(false),
})
if err != nil {
return err
}
subnet, err := compute.NewSubnetwork(ctx, "subnet", &compute.SubnetworkArgs{
Name: pulumi.String("mdc-subnet"),
IpCidrRange: pulumi.String("10.0.0.0/24"),
Region: pulumi.String("us-east4"),
Network: net.ID(),
PrivateIpGoogleAccess: pulumi.Bool(true),
})
if err != nil {
return err
}
internal, err := compute.NewFirewall(ctx, "internal", &compute.FirewallArgs{
Name: pulumi.String("mdc-fw"),
Network: net.Name,
Direction: pulumi.String("INGRESS"),
SourceRanges: pulumi.StringArray{
pulumi.String("10.0.0.0/24"),
},
Allows: compute.FirewallAllowArray{
&compute.FirewallAllowArgs{
Protocol: pulumi.String("tcp"),
Ports: pulumi.StringArray{
pulumi.String("0-65535"),
},
},
&compute.FirewallAllowArgs{
Protocol: pulumi.String("udp"),
Ports: pulumi.StringArray{
pulumi.String("0-65535"),
},
},
&compute.FirewallAllowArgs{
Protocol: pulumi.String("icmp"),
},
},
})
if err != nil {
return err
}
home, err := filestore.NewInstance(ctx, "home", &filestore.InstanceArgs{
Name: pulumi.String("mdc-home"),
Location: pulumi.String("us-east4-b"),
Tier: pulumi.String("BASIC_HDD"),
FileShares: &filestore.InstanceFileSharesArgs{
CapacityGb: pulumi.Int(1024),
Name: pulumi.String("share1"),
},
Networks: filestore.InstanceNetworkArray{
&filestore.InstanceNetworkArgs{
Network: net.Name,
Modes: pulumi.StringArray{
pulumi.String("MODE_IPV4"),
},
},
},
})
if err != nil {
return err
}
data, err := filestore.NewInstance(ctx, "data", &filestore.InstanceArgs{
Name: pulumi.String("mdc-data"),
Location: pulumi.String("us-east4-b"),
Tier: pulumi.String("BASIC_HDD"),
FileShares: &filestore.InstanceFileSharesArgs{
CapacityGb: pulumi.Int(1024),
Name: pulumi.String("share1"),
},
Networks: filestore.InstanceNetworkArray{
&filestore.InstanceNetworkArgs{
Network: net.Name,
Modes: pulumi.StringArray{
pulumi.String("MODE_IPV4"),
},
},
},
})
if err != nil {
return err
}
res, err := compute.NewReservation(ctx, "res", &compute.ReservationArgs{
Name: pulumi.String("mdc-res"),
Zone: pulumi.String("us-east4-b"),
SpecificReservationRequired: pulumi.Bool(true),
SpecificReservation: &compute.ReservationSpecificReservationArgs{
Count: pulumi.Int(1),
InstanceProperties: &compute.ReservationSpecificReservationInstancePropertiesArgs{
MachineType: pulumi.String("n2-standard-4"),
},
},
})
if err != nil {
return err
}
_, err = vertex.NewAiModelDevelopmentCluster(ctx, "default", &vertex.AiModelDevelopmentClusterArgs{
Region: pulumi.String("us-east4"),
ModelDevelopmentClusterId: pulumi.String("tftest"),
DisplayName: pulumi.String("tftest"),
DeletionPolicy: pulumi.String("DELETE"),
UpdateMode: pulumi.String("USER_AND_SERVICE"),
Labels: pulumi.StringMap{
"env": pulumi.String("dev"),
},
RuntimeSpec: &vertex.AiModelDevelopmentClusterRuntimeSpecArgs{
ServiceAccount: pulumi.Sprintf("%v-compute@developer.gserviceaccount.com", project.Number),
},
LustreExportConfig: &vertex.AiModelDevelopmentClusterLustreExportConfigArgs{
DestinationBucket: pulumi.String("mdc-backup-bucket"),
DaysOfWeeks: pulumi.StringArray{
pulumi.String("MONDAY"),
pulumi.String("THURSDAY"),
},
StartTime: &vertex.AiModelDevelopmentClusterLustreExportConfigStartTimeArgs{
Hours: pulumi.Int(4),
Minutes: pulumi.Int(30),
},
},
Network: &vertex.AiModelDevelopmentClusterNetworkArgs{
Network: net.ID(),
Subnetwork: subnet.ID(),
},
OrchestratorSpec: &vertex.AiModelDevelopmentClusterOrchestratorSpecArgs{
SlurmSpec: &vertex.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecArgs{
HomeDirectoryStorage: home.ID(),
LoginNodePoolId: pulumi.String("login"),
DefaultPartitionId: pulumi.String("main"),
PrologBashScripts: pulumi.StringArray{
pulumi.String("#!/bin/bash\necho prolog"),
},
EpilogBashScripts: pulumi.StringArray{
pulumi.String("#!/bin/bash\necho epilog"),
},
Config: pulumi.StringMap{
"PriorityType": pulumi.String("priority/multifactor"),
"PriorityWeightAge": pulumi.String("1000"),
},
NodeSets: vertex.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecNodeSetArray{
&vertex.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecNodeSetArgs{
NodePoolId: pulumi.String("compute"),
Config: pulumi.StringMap{
"Weight": pulumi.String("10"),
},
},
},
Partitions: vertex.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecPartitionArray{
&vertex.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecPartitionArgs{
Id: pulumi.String("main"),
NodePoolIds: pulumi.StringArray{
pulumi.String("compute"),
pulumi.String("flexpool"),
pulumi.String("respool"),
},
Config: pulumi.StringMap{
"PriorityTier": pulumi.String("10"),
},
},
},
},
},
NodePools: vertex.AiModelDevelopmentClusterNodePoolArray{
&vertex.AiModelDevelopmentClusterNodePoolArgs{
Id: pulumi.String("compute"),
Zone: pulumi.String("us-east4-b"),
ProvisioningModel: pulumi.String("ON_DEMAND"),
NodeImage: pulumi.String("projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113"),
StartupScript: pulumi.String("#!/bin/bash\necho hello"),
EnablePublicIps: pulumi.Bool(true),
Filestores: pulumi.StringArray{
data.ID(),
},
Labels: pulumi.StringMap{
"pool": pulumi.String("compute"),
},
BootDisk: &vertex.AiModelDevelopmentClusterNodePoolBootDiskArgs{
BootDiskSizeGb: pulumi.Int(100),
BootDiskType: pulumi.String("pd-ssd"),
},
MachineSpec: &vertex.AiModelDevelopmentClusterNodePoolMachineSpecArgs{
MachineType: pulumi.String("n2-standard-4"),
},
ScalingSpec: &vertex.AiModelDevelopmentClusterNodePoolScalingSpecArgs{
MinNodeCount: pulumi.Int(1),
MaxNodeCount: pulumi.Int(2),
},
},
&vertex.AiModelDevelopmentClusterNodePoolArgs{
Id: pulumi.String("flexpool"),
Zone: pulumi.String("us-east4-b"),
ProvisioningModel: pulumi.String("FLEX_START"),
FlexStartMaxDuration: pulumi.String("3600s"),
NodeImage: pulumi.String("projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113"),
StartupScript: pulumi.String("#!/bin/bash\necho hello"),
EnablePublicIps: pulumi.Bool(true),
Filestores: pulumi.StringArray{
data.ID(),
},
Labels: pulumi.StringMap{
"pool": pulumi.String("flexpool"),
},
BootDisk: &vertex.AiModelDevelopmentClusterNodePoolBootDiskArgs{
BootDiskSizeGb: pulumi.Int(100),
BootDiskType: pulumi.String("hyperdisk-balanced"),
},
MachineSpec: &vertex.AiModelDevelopmentClusterNodePoolMachineSpecArgs{
MachineType: pulumi.String("a3-ultragpu-8g"),
},
ScalingSpec: &vertex.AiModelDevelopmentClusterNodePoolScalingSpecArgs{
MinNodeCount: pulumi.Int(0),
MaxNodeCount: pulumi.Int(1),
},
},
&vertex.AiModelDevelopmentClusterNodePoolArgs{
Id: pulumi.String("login"),
Zone: pulumi.String("us-east4-b"),
ProvisioningModel: pulumi.String("ON_DEMAND"),
NodeImage: pulumi.String("projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113"),
StartupScript: pulumi.String("#!/bin/bash\necho hello"),
EnablePublicIps: pulumi.Bool(true),
Filestores: pulumi.StringArray{
data.ID(),
},
Labels: pulumi.StringMap{
"pool": pulumi.String("login"),
},
BootDisk: &vertex.AiModelDevelopmentClusterNodePoolBootDiskArgs{
BootDiskSizeGb: pulumi.Int(100),
BootDiskType: pulumi.String("pd-ssd"),
},
MachineSpec: &vertex.AiModelDevelopmentClusterNodePoolMachineSpecArgs{
MachineType: pulumi.String("n2-standard-4"),
},
ScalingSpec: &vertex.AiModelDevelopmentClusterNodePoolScalingSpecArgs{
MinNodeCount: pulumi.Int(1),
MaxNodeCount: pulumi.Int(1),
},
},
&vertex.AiModelDevelopmentClusterNodePoolArgs{
Id: pulumi.String("respool"),
Zone: pulumi.String("us-east4-b"),
ProvisioningModel: pulumi.String("RESERVATION"),
NodeImage: pulumi.String("projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113"),
StartupScript: pulumi.String("#!/bin/bash\necho hello"),
EnablePublicIps: pulumi.Bool(true),
Filestores: pulumi.StringArray{
data.ID(),
},
Labels: pulumi.StringMap{
"pool": pulumi.String("respool"),
},
BootDisk: &vertex.AiModelDevelopmentClusterNodePoolBootDiskArgs{
BootDiskSizeGb: pulumi.Int(100),
BootDiskType: pulumi.String("pd-ssd"),
},
MachineSpec: &vertex.AiModelDevelopmentClusterNodePoolMachineSpecArgs{
MachineType: pulumi.String("n2-standard-4"),
ReservationAffinity: &vertex.AiModelDevelopmentClusterNodePoolMachineSpecReservationAffinityArgs{
ReservationAffinityType: pulumi.String("SPECIFIC_RESERVATION"),
Values: pulumi.StringArray{
res.ID(),
},
},
},
ScalingSpec: &vertex.AiModelDevelopmentClusterNodePoolScalingSpecArgs{
MinNodeCount: pulumi.Int(0),
MaxNodeCount: pulumi.Int(1),
},
},
},
}, pulumi.DependsOn([]pulumi.Resource{
internal,
}))
if err != nil {
return err
}
return nil
})
}
using System.Collections.Generic;
using System.Linq;
using Pulumi;
using Gcp = Pulumi.Gcp;
return await Deployment.RunAsync(() =>
{
var project = Gcp.Organizations.GetProject.Invoke();
var net = new Gcp.Compute.Network("net", new()
{
Name = "mdc-net",
AutoCreateSubnetworks = false,
});
var subnet = new Gcp.Compute.Subnetwork("subnet", new()
{
Name = "mdc-subnet",
IpCidrRange = "10.0.0.0/24",
Region = "us-east4",
Network = net.Id,
PrivateIpGoogleAccess = true,
});
var @internal = new Gcp.Compute.Firewall("internal", new()
{
Name = "mdc-fw",
Network = net.Name,
Direction = "INGRESS",
SourceRanges = new[]
{
"10.0.0.0/24",
},
Allows = new[]
{
new Gcp.Compute.Inputs.FirewallAllowArgs
{
Protocol = "tcp",
Ports = new[]
{
"0-65535",
},
},
new Gcp.Compute.Inputs.FirewallAllowArgs
{
Protocol = "udp",
Ports = new[]
{
"0-65535",
},
},
new Gcp.Compute.Inputs.FirewallAllowArgs
{
Protocol = "icmp",
},
},
});
var home = new Gcp.Filestore.Instance("home", new()
{
Name = "mdc-home",
Location = "us-east4-b",
Tier = "BASIC_HDD",
FileShares = new Gcp.Filestore.Inputs.InstanceFileSharesArgs
{
CapacityGb = 1024,
Name = "share1",
},
Networks = new[]
{
new Gcp.Filestore.Inputs.InstanceNetworkArgs
{
Network = net.Name,
Modes = new[]
{
"MODE_IPV4",
},
},
},
});
var data = new Gcp.Filestore.Instance("data", new()
{
Name = "mdc-data",
Location = "us-east4-b",
Tier = "BASIC_HDD",
FileShares = new Gcp.Filestore.Inputs.InstanceFileSharesArgs
{
CapacityGb = 1024,
Name = "share1",
},
Networks = new[]
{
new Gcp.Filestore.Inputs.InstanceNetworkArgs
{
Network = net.Name,
Modes = new[]
{
"MODE_IPV4",
},
},
},
});
var res = new Gcp.Compute.Reservation("res", new()
{
Name = "mdc-res",
Zone = "us-east4-b",
SpecificReservationRequired = true,
SpecificReservation = new Gcp.Compute.Inputs.ReservationSpecificReservationArgs
{
Count = 1,
InstanceProperties = new Gcp.Compute.Inputs.ReservationSpecificReservationInstancePropertiesArgs
{
MachineType = "n2-standard-4",
},
},
});
var @default = new Gcp.Vertex.AiModelDevelopmentCluster("default", new()
{
Region = "us-east4",
ModelDevelopmentClusterId = "tftest",
DisplayName = "tftest",
DeletionPolicy = "DELETE",
UpdateMode = "USER_AND_SERVICE",
Labels =
{
{ "env", "dev" },
},
RuntimeSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterRuntimeSpecArgs
{
ServiceAccount = $"{project.Apply(getProjectResult => getProjectResult.Number)}-compute@developer.gserviceaccount.com",
},
LustreExportConfig = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterLustreExportConfigArgs
{
DestinationBucket = "mdc-backup-bucket",
DaysOfWeeks = new[]
{
"MONDAY",
"THURSDAY",
},
StartTime = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterLustreExportConfigStartTimeArgs
{
Hours = 4,
Minutes = 30,
},
},
Network = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNetworkArgs
{
Network = net.Id,
Subnetwork = subnet.Id,
},
OrchestratorSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterOrchestratorSpecArgs
{
SlurmSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecArgs
{
HomeDirectoryStorage = home.Id,
LoginNodePoolId = "login",
DefaultPartitionId = "main",
PrologBashScripts = new[]
{
@"#!/bin/bash
echo prolog",
},
EpilogBashScripts = new[]
{
@"#!/bin/bash
echo epilog",
},
Config =
{
{ "PriorityType", "priority/multifactor" },
{ "PriorityWeightAge", "1000" },
},
NodeSets = new[]
{
new Gcp.Vertex.Inputs.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecNodeSetArgs
{
NodePoolId = "compute",
Config =
{
{ "Weight", "10" },
},
},
},
Partitions = new[]
{
new Gcp.Vertex.Inputs.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecPartitionArgs
{
Id = "main",
NodePoolIds = new[]
{
"compute",
"flexpool",
"respool",
},
Config =
{
{ "PriorityTier", "10" },
},
},
},
},
},
NodePools = new[]
{
new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolArgs
{
Id = "compute",
Zone = "us-east4-b",
ProvisioningModel = "ON_DEMAND",
NodeImage = "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113",
StartupScript = @"#!/bin/bash
echo hello",
EnablePublicIps = true,
Filestores = new[]
{
data.Id,
},
Labels =
{
{ "pool", "compute" },
},
BootDisk = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolBootDiskArgs
{
BootDiskSizeGb = 100,
BootDiskType = "pd-ssd",
},
MachineSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolMachineSpecArgs
{
MachineType = "n2-standard-4",
},
ScalingSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolScalingSpecArgs
{
MinNodeCount = 1,
MaxNodeCount = 2,
},
},
new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolArgs
{
Id = "flexpool",
Zone = "us-east4-b",
ProvisioningModel = "FLEX_START",
FlexStartMaxDuration = "3600s",
NodeImage = "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113",
StartupScript = @"#!/bin/bash
echo hello",
EnablePublicIps = true,
Filestores = new[]
{
data.Id,
},
Labels =
{
{ "pool", "flexpool" },
},
BootDisk = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolBootDiskArgs
{
BootDiskSizeGb = 100,
BootDiskType = "hyperdisk-balanced",
},
MachineSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolMachineSpecArgs
{
MachineType = "a3-ultragpu-8g",
},
ScalingSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolScalingSpecArgs
{
MinNodeCount = 0,
MaxNodeCount = 1,
},
},
new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolArgs
{
Id = "login",
Zone = "us-east4-b",
ProvisioningModel = "ON_DEMAND",
NodeImage = "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113",
StartupScript = @"#!/bin/bash
echo hello",
EnablePublicIps = true,
Filestores = new[]
{
data.Id,
},
Labels =
{
{ "pool", "login" },
},
BootDisk = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolBootDiskArgs
{
BootDiskSizeGb = 100,
BootDiskType = "pd-ssd",
},
MachineSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolMachineSpecArgs
{
MachineType = "n2-standard-4",
},
ScalingSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolScalingSpecArgs
{
MinNodeCount = 1,
MaxNodeCount = 1,
},
},
new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolArgs
{
Id = "respool",
Zone = "us-east4-b",
ProvisioningModel = "RESERVATION",
NodeImage = "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113",
StartupScript = @"#!/bin/bash
echo hello",
EnablePublicIps = true,
Filestores = new[]
{
data.Id,
},
Labels =
{
{ "pool", "respool" },
},
BootDisk = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolBootDiskArgs
{
BootDiskSizeGb = 100,
BootDiskType = "pd-ssd",
},
MachineSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolMachineSpecArgs
{
MachineType = "n2-standard-4",
ReservationAffinity = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolMachineSpecReservationAffinityArgs
{
ReservationAffinityType = "SPECIFIC_RESERVATION",
Values = new[]
{
res.Id,
},
},
},
ScalingSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolScalingSpecArgs
{
MinNodeCount = 0,
MaxNodeCount = 1,
},
},
},
}, new CustomResourceOptions
{
DependsOn =
{
@internal,
},
});
});
package generated_program;
import com.pulumi.Context;
import com.pulumi.Pulumi;
import com.pulumi.core.Output;
import com.pulumi.gcp.organizations.OrganizationsFunctions;
import com.pulumi.gcp.organizations.inputs.GetProjectArgs;
import com.pulumi.gcp.compute.Network;
import com.pulumi.gcp.compute.NetworkArgs;
import com.pulumi.gcp.compute.Subnetwork;
import com.pulumi.gcp.compute.SubnetworkArgs;
import com.pulumi.gcp.compute.Firewall;
import com.pulumi.gcp.compute.FirewallArgs;
import com.pulumi.gcp.compute.inputs.FirewallAllowArgs;
import com.pulumi.gcp.filestore.Instance;
import com.pulumi.gcp.filestore.InstanceArgs;
import com.pulumi.gcp.filestore.inputs.InstanceFileSharesArgs;
import com.pulumi.gcp.filestore.inputs.InstanceNetworkArgs;
import com.pulumi.gcp.compute.Reservation;
import com.pulumi.gcp.compute.ReservationArgs;
import com.pulumi.gcp.compute.inputs.ReservationSpecificReservationArgs;
import com.pulumi.gcp.compute.inputs.ReservationSpecificReservationInstancePropertiesArgs;
import com.pulumi.gcp.vertex.AiModelDevelopmentCluster;
import com.pulumi.gcp.vertex.AiModelDevelopmentClusterArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterRuntimeSpecArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterLustreExportConfigArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterLustreExportConfigStartTimeArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterNetworkArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterOrchestratorSpecArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecNodeSetArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecPartitionArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterNodePoolArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterNodePoolBootDiskArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterNodePoolMachineSpecArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterNodePoolScalingSpecArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterNodePoolMachineSpecReservationAffinityArgs;
import com.pulumi.resources.CustomResourceOptions;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.Map;
import java.io.File;
import java.nio.file.Files;
import java.nio.file.Paths;
public class App {
public static void main(String[] args) {
Pulumi.run(App::stack);
}
public static void stack(Context ctx) {
final var project = OrganizationsFunctions.getProject(GetProjectArgs.builder()
.build());
var net = new Network("net", NetworkArgs.builder()
.name("mdc-net")
.autoCreateSubnetworks(false)
.build());
var subnet = new Subnetwork("subnet", SubnetworkArgs.builder()
.name("mdc-subnet")
.ipCidrRange("10.0.0.0/24")
.region("us-east4")
.network(net.id())
.privateIpGoogleAccess(true)
.build());
var internal = new Firewall("internal", FirewallArgs.builder()
.name("mdc-fw")
.network(net.name())
.direction("INGRESS")
.sourceRanges("10.0.0.0/24")
.allows(
FirewallAllowArgs.builder()
.protocol("tcp")
.ports("0-65535")
.build(),
FirewallAllowArgs.builder()
.protocol("udp")
.ports("0-65535")
.build(),
FirewallAllowArgs.builder()
.protocol("icmp")
.build())
.build());
var home = new Instance("home", InstanceArgs.builder()
.name("mdc-home")
.location("us-east4-b")
.tier("BASIC_HDD")
.fileShares(InstanceFileSharesArgs.builder()
.capacityGb(1024)
.name("share1")
.build())
.networks(InstanceNetworkArgs.builder()
.network(net.name())
.modes("MODE_IPV4")
.build())
.build());
var data = new Instance("data", InstanceArgs.builder()
.name("mdc-data")
.location("us-east4-b")
.tier("BASIC_HDD")
.fileShares(InstanceFileSharesArgs.builder()
.capacityGb(1024)
.name("share1")
.build())
.networks(InstanceNetworkArgs.builder()
.network(net.name())
.modes("MODE_IPV4")
.build())
.build());
var res = new Reservation("res", ReservationArgs.builder()
.name("mdc-res")
.zone("us-east4-b")
.specificReservationRequired(true)
.specificReservation(ReservationSpecificReservationArgs.builder()
.count(1)
.instanceProperties(ReservationSpecificReservationInstancePropertiesArgs.builder()
.machineType("n2-standard-4")
.build())
.build())
.build());
var default_ = new AiModelDevelopmentCluster("default", AiModelDevelopmentClusterArgs.builder()
.region("us-east4")
.modelDevelopmentClusterId("tftest")
.displayName("tftest")
.deletionPolicy("DELETE")
.updateMode("USER_AND_SERVICE")
.labels(Map.of("env", "dev"))
.runtimeSpec(AiModelDevelopmentClusterRuntimeSpecArgs.builder()
.serviceAccount(String.format("%s-compute@developer.gserviceaccount.com", project.number()))
.build())
.lustreExportConfig(AiModelDevelopmentClusterLustreExportConfigArgs.builder()
.destinationBucket("mdc-backup-bucket")
.daysOfWeeks(
"MONDAY",
"THURSDAY")
.startTime(AiModelDevelopmentClusterLustreExportConfigStartTimeArgs.builder()
.hours(4)
.minutes(30)
.build())
.build())
.network(AiModelDevelopmentClusterNetworkArgs.builder()
.network(net.id())
.subnetwork(subnet.id())
.build())
.orchestratorSpec(AiModelDevelopmentClusterOrchestratorSpecArgs.builder()
.slurmSpec(AiModelDevelopmentClusterOrchestratorSpecSlurmSpecArgs.builder()
.homeDirectoryStorage(home.id())
.loginNodePoolId("login")
.defaultPartitionId("main")
.prologBashScripts("""
#!/bin/bash
echo prolog """)
.epilogBashScripts("""
#!/bin/bash
echo epilog """)
.config(Map.ofEntries(
Map.entry("PriorityType", "priority/multifactor"),
Map.entry("PriorityWeightAge", "1000")
))
.nodeSets(AiModelDevelopmentClusterOrchestratorSpecSlurmSpecNodeSetArgs.builder()
.nodePoolId("compute")
.config(Map.of("Weight", "10"))
.build())
.partitions(AiModelDevelopmentClusterOrchestratorSpecSlurmSpecPartitionArgs.builder()
.id("main")
.nodePoolIds(
"compute",
"flexpool",
"respool")
.config(Map.of("PriorityTier", "10"))
.build())
.build())
.build())
.nodePools(
AiModelDevelopmentClusterNodePoolArgs.builder()
.id("compute")
.zone("us-east4-b")
.provisioningModel("ON_DEMAND")
.nodeImage("projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113")
.startupScript("""
#!/bin/bash
echo hello """)
.enablePublicIps(true)
.filestores(data.id())
.labels(Map.of("pool", "compute"))
.bootDisk(AiModelDevelopmentClusterNodePoolBootDiskArgs.builder()
.bootDiskSizeGb(100)
.bootDiskType("pd-ssd")
.build())
.machineSpec(AiModelDevelopmentClusterNodePoolMachineSpecArgs.builder()
.machineType("n2-standard-4")
.build())
.scalingSpec(AiModelDevelopmentClusterNodePoolScalingSpecArgs.builder()
.minNodeCount(1)
.maxNodeCount(2)
.build())
.build(),
AiModelDevelopmentClusterNodePoolArgs.builder()
.id("flexpool")
.zone("us-east4-b")
.provisioningModel("FLEX_START")
.flexStartMaxDuration("3600s")
.nodeImage("projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113")
.startupScript("""
#!/bin/bash
echo hello """)
.enablePublicIps(true)
.filestores(data.id())
.labels(Map.of("pool", "flexpool"))
.bootDisk(AiModelDevelopmentClusterNodePoolBootDiskArgs.builder()
.bootDiskSizeGb(100)
.bootDiskType("hyperdisk-balanced")
.build())
.machineSpec(AiModelDevelopmentClusterNodePoolMachineSpecArgs.builder()
.machineType("a3-ultragpu-8g")
.build())
.scalingSpec(AiModelDevelopmentClusterNodePoolScalingSpecArgs.builder()
.minNodeCount(0)
.maxNodeCount(1)
.build())
.build(),
AiModelDevelopmentClusterNodePoolArgs.builder()
.id("login")
.zone("us-east4-b")
.provisioningModel("ON_DEMAND")
.nodeImage("projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113")
.startupScript("""
#!/bin/bash
echo hello """)
.enablePublicIps(true)
.filestores(data.id())
.labels(Map.of("pool", "login"))
.bootDisk(AiModelDevelopmentClusterNodePoolBootDiskArgs.builder()
.bootDiskSizeGb(100)
.bootDiskType("pd-ssd")
.build())
.machineSpec(AiModelDevelopmentClusterNodePoolMachineSpecArgs.builder()
.machineType("n2-standard-4")
.build())
.scalingSpec(AiModelDevelopmentClusterNodePoolScalingSpecArgs.builder()
.minNodeCount(1)
.maxNodeCount(1)
.build())
.build(),
AiModelDevelopmentClusterNodePoolArgs.builder()
.id("respool")
.zone("us-east4-b")
.provisioningModel("RESERVATION")
.nodeImage("projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113")
.startupScript("""
#!/bin/bash
echo hello """)
.enablePublicIps(true)
.filestores(data.id())
.labels(Map.of("pool", "respool"))
.bootDisk(AiModelDevelopmentClusterNodePoolBootDiskArgs.builder()
.bootDiskSizeGb(100)
.bootDiskType("pd-ssd")
.build())
.machineSpec(AiModelDevelopmentClusterNodePoolMachineSpecArgs.builder()
.machineType("n2-standard-4")
.reservationAffinity(AiModelDevelopmentClusterNodePoolMachineSpecReservationAffinityArgs.builder()
.reservationAffinityType("SPECIFIC_RESERVATION")
.values(res.id())
.build())
.build())
.scalingSpec(AiModelDevelopmentClusterNodePoolScalingSpecArgs.builder()
.minNodeCount(0)
.maxNodeCount(1)
.build())
.build())
.build(), CustomResourceOptions.builder()
.dependsOn(internal)
.build());
}
}
resources:
net:
type: gcp:compute:Network
properties:
name: mdc-net
autoCreateSubnetworks: false
subnet:
type: gcp:compute:Subnetwork
properties:
name: mdc-subnet
ipCidrRange: 10.0.0.0/24
region: us-east4
network: ${net.id}
privateIpGoogleAccess: true
internal:
type: gcp:compute:Firewall
properties:
name: mdc-fw
network: ${net.name}
direction: INGRESS
sourceRanges:
- 10.0.0.0/24
allows:
- protocol: tcp
ports:
- 0-65535
- protocol: udp
ports:
- 0-65535
- protocol: icmp
home:
type: gcp:filestore:Instance
properties:
name: mdc-home
location: us-east4-b
tier: BASIC_HDD
fileShares:
capacityGb: 1024
name: share1
networks:
- network: ${net.name}
modes:
- MODE_IPV4
data:
type: gcp:filestore:Instance
properties:
name: mdc-data
location: us-east4-b
tier: BASIC_HDD
fileShares:
capacityGb: 1024
name: share1
networks:
- network: ${net.name}
modes:
- MODE_IPV4
res:
type: gcp:compute:Reservation
properties:
name: mdc-res
zone: us-east4-b
specificReservationRequired: true
specificReservation:
count: 1
instanceProperties:
machineType: n2-standard-4
default:
type: gcp:vertex:AiModelDevelopmentCluster
properties:
region: us-east4
modelDevelopmentClusterId: tftest
displayName: tftest
deletionPolicy: DELETE
updateMode: USER_AND_SERVICE
labels:
env: dev
runtimeSpec:
serviceAccount: ${project.number}-compute@developer.gserviceaccount.com
lustreExportConfig:
destinationBucket: mdc-backup-bucket
daysOfWeeks:
- MONDAY
- THURSDAY
startTime:
hours: 4
minutes: 30
network:
network: ${net.id}
subnetwork: ${subnet.id}
orchestratorSpec:
slurmSpec:
homeDirectoryStorage: ${home.id}
loginNodePoolId: login
defaultPartitionId: main
prologBashScripts:
- |-
#!/bin/bash
echo prolog
epilogBashScripts:
- |-
#!/bin/bash
echo epilog
config:
PriorityType: priority/multifactor
PriorityWeightAge: '1000'
nodeSets:
- nodePoolId: compute
config:
Weight: '10'
partitions:
- id: main
nodePoolIds:
- compute
- flexpool
- respool
config:
PriorityTier: '10'
nodePools:
- id: compute
zone: us-east4-b
provisioningModel: ON_DEMAND
nodeImage: projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113
startupScript: |-
#!/bin/bash
echo hello
enablePublicIps: true
filestores:
- ${data.id}
labels:
pool: compute
bootDisk:
bootDiskSizeGb: 100
bootDiskType: pd-ssd
machineSpec:
machineType: n2-standard-4
scalingSpec:
minNodeCount: 1
maxNodeCount: 2
- id: flexpool
zone: us-east4-b
provisioningModel: FLEX_START
flexStartMaxDuration: 3600s
nodeImage: projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113
startupScript: |-
#!/bin/bash
echo hello
enablePublicIps: true
filestores:
- ${data.id}
labels:
pool: flexpool
bootDisk:
bootDiskSizeGb: 100
bootDiskType: hyperdisk-balanced
machineSpec:
machineType: a3-ultragpu-8g
scalingSpec:
minNodeCount: 0
maxNodeCount: 1
- id: login
zone: us-east4-b
provisioningModel: ON_DEMAND
nodeImage: projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113
startupScript: |-
#!/bin/bash
echo hello
enablePublicIps: true
filestores:
- ${data.id}
labels:
pool: login
bootDisk:
bootDiskSizeGb: 100
bootDiskType: pd-ssd
machineSpec:
machineType: n2-standard-4
scalingSpec:
minNodeCount: 1
maxNodeCount: 1
- id: respool
zone: us-east4-b
provisioningModel: RESERVATION
nodeImage: projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113
startupScript: |-
#!/bin/bash
echo hello
enablePublicIps: true
filestores:
- ${data.id}
labels:
pool: respool
bootDisk:
bootDiskSizeGb: 100
bootDiskType: pd-ssd
machineSpec:
machineType: n2-standard-4
reservationAffinity:
reservationAffinityType: SPECIFIC_RESERVATION
values:
- ${res.id}
scalingSpec:
minNodeCount: 0
maxNodeCount: 1
options:
dependsOn:
- ${internal}
variables:
project:
fn::invoke:
function: gcp:organizations:getProject
arguments: {}
terraform {
required_providers {
gcp = {
source = "pulumi/gcp"
}
}
}
data "gcp_organizations_project" "project" {
}
resource "gcp_compute_network" "net" {
lifecycle {
create_before_destroy = true
}
name = "mdc-net"
auto_create_subnetworks = false
}
resource "gcp_compute_subnetwork" "subnet" {
lifecycle {
create_before_destroy = true
}
name = "mdc-subnet"
ip_cidr_range = "10.0.0.0/24"
region = "us-east4"
network = gcp_compute_network.net.id
private_ip_google_access = true
}
resource "gcp_compute_firewall" "internal" {
lifecycle {
create_before_destroy = true
}
name = "mdc-fw"
network = gcp_compute_network.net.name
direction = "INGRESS"
source_ranges = ["10.0.0.0/24"]
allows {
protocol = "tcp"
ports = ["0-65535"]
}
allows {
protocol = "udp"
ports = ["0-65535"]
}
allows {
protocol = "icmp"
}
}
resource "gcp_filestore_instance" "home" {
lifecycle {
create_before_destroy = true
}
name = "mdc-home"
location = "us-east4-b"
tier = "BASIC_HDD"
file_shares = {
capacity_gb = 1024
name = "share1"
}
networks {
network = gcp_compute_network.net.name
modes = ["MODE_IPV4"]
}
}
resource "gcp_filestore_instance" "data" {
lifecycle {
create_before_destroy = true
}
name = "mdc-data"
location = "us-east4-b"
tier = "BASIC_HDD"
file_shares = {
capacity_gb = 1024
name = "share1"
}
networks {
network = gcp_compute_network.net.name
modes = ["MODE_IPV4"]
}
}
resource "gcp_compute_reservation" "res" {
lifecycle {
create_before_destroy = true
}
name = "mdc-res"
zone = "us-east4-b"
specific_reservation_required = true
specific_reservation = {
count = 1
instance_properties = {
machine_type = "n2-standard-4"
}
}
}
resource "gcp_vertex_ai_model_development_cluster" "default" {
depends_on = [gcp_compute_firewall.internal]
lifecycle {
create_before_destroy = true
}
region = "us-east4"
model_development_cluster_id = "tftest"
display_name = "tftest"
deletion_policy = "DELETE"
update_mode = "USER_AND_SERVICE"
labels = {
"env" = "dev"
}
runtime_spec = {
service_account ="${data.gcp_organizations_project.project.number}-compute@developer.gserviceaccount.com"
}
lustre_export_config = {
destination_bucket = "mdc-backup-bucket"
days_of_weeks = ["MONDAY", "THURSDAY"]
start_time = {
hours = 4
minutes = 30
}
}
network = {
network = gcp_compute_network.net.id
subnetwork = gcp_compute_subnetwork.subnet.id
}
orchestrator_spec = {
slurm_spec = {
home_directory_storage = gcp_filestore_instance.home.id
login_node_pool_id = "login"
default_partition_id = "main"
prolog_bash_scripts = ["#!/bin/bash\necho prolog"]
epilog_bash_scripts = ["#!/bin/bash\necho epilog"]
config = {
"PriorityType" = "priority/multifactor"
"PriorityWeightAge" = "1000"
}
node_sets = [{
node_pool_id = "compute"
config = {
"Weight" = "10"
}
}]
partitions = [{
id = "main"
node_pool_ids = ["compute", "flexpool", "respool"]
config = {
"PriorityTier" = "10"
}
}]
}
}
node_pools {
id = "compute"
zone = "us-east4-b"
provisioning_model = "ON_DEMAND"
node_image = "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113"
startup_script = "#!/bin/bash\necho hello"
enable_public_ips = true
filestores = [gcp_filestore_instance.data.id]
labels = {
"pool" = "compute"
}
boot_disk = {
boot_disk_size_gb = 100
boot_disk_type = "pd-ssd"
}
machine_spec = {
machine_type = "n2-standard-4"
}
scaling_spec = {
min_node_count = 1
max_node_count = 2
}
}
node_pools {
id = "flexpool"
zone = "us-east4-b"
provisioning_model = "FLEX_START"
flex_start_max_duration = "3600s"
node_image = "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113"
startup_script = "#!/bin/bash\necho hello"
enable_public_ips = true
filestores = [gcp_filestore_instance.data.id]
labels = {
"pool" = "flexpool"
}
boot_disk = {
boot_disk_size_gb = 100
boot_disk_type = "hyperdisk-balanced"
}
machine_spec = {
machine_type = "a3-ultragpu-8g"
}
scaling_spec = {
min_node_count = 0
max_node_count = 1
}
}
node_pools {
id = "login"
zone = "us-east4-b"
provisioning_model = "ON_DEMAND"
node_image = "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113"
startup_script = "#!/bin/bash\necho hello"
enable_public_ips = true
filestores = [gcp_filestore_instance.data.id]
labels = {
"pool" = "login"
}
boot_disk = {
boot_disk_size_gb = 100
boot_disk_type = "pd-ssd"
}
machine_spec = {
machine_type = "n2-standard-4"
}
scaling_spec = {
min_node_count = 1
max_node_count = 1
}
}
node_pools {
id = "respool"
zone = "us-east4-b"
provisioning_model = "RESERVATION"
node_image = "projects/clusterdirector-public-images/global/images/family/common-ubuntu-2204-amd64-nvidia-580-slurm-2505-v20251113"
startup_script = "#!/bin/bash\necho hello"
enable_public_ips = true
filestores = [gcp_filestore_instance.data.id]
labels = {
"pool" = "respool"
}
boot_disk = {
boot_disk_size_gb = 100
boot_disk_type = "pd-ssd"
}
machine_spec = {
machine_type = "n2-standard-4"
reservation_affinity = {
reservation_affinity_type = "SPECIFIC_RESERVATION"
values = [gcp_compute_reservation.res.id]
}
}
scaling_spec = {
min_node_count = 0
max_node_count = 1
}
}
}
Vertex Ai Model Development Cluster Gke
import * as pulumi from "@pulumi/pulumi";
import * as gcp from "@pulumi/gcp";
const net = new gcp.compute.Network("net", {
name: "mdc-gke-net",
autoCreateSubnetworks: false,
});
const subnet = new gcp.compute.Subnetwork("subnet", {
name: "mdc-gke-subnet",
ipCidrRange: "10.0.0.0/24",
region: "us-east1",
network: net.id,
privateIpGoogleAccess: true,
});
const internal = new gcp.compute.Firewall("internal", {
name: "mdc-gke-fw",
network: net.name,
direction: "INGRESS",
sourceRanges: ["10.0.0.0/24"],
allows: [
{
protocol: "tcp",
ports: ["0-65535"],
},
{
protocol: "udp",
ports: ["0-65535"],
},
{
protocol: "icmp",
},
],
});
const _default = new gcp.vertex.AiModelDevelopmentCluster("default", {
region: "us-east1",
modelDevelopmentClusterId: "tftest-gke",
displayName: "tftest-gke",
deletionPolicy: "DELETE",
labels: {
env: "dev",
},
network: {
network: net.id,
subnetwork: subnet.id,
},
orchestratorSpec: {
kubernetesSpec: {
nodePoolIds: ["cpu1"],
enableIpAlias: true,
publicAccessEndpointEnabled: true,
controlPlaneVersion: "-",
nodePoolVersions: {
cpu1: "-",
},
},
},
nodePools: [{
id: "cpu1",
zone: "us-east1-b",
provisioningModel: "ON_DEMAND",
bootDisk: {
bootDiskSizeGb: 100,
bootDiskType: "pd-standard",
},
machineSpec: {
machineType: "n2-standard-4",
},
scalingSpec: {
minNodeCount: 1,
maxNodeCount: 2,
},
}],
}, {
dependsOn: [internal],
});
import pulumi
import pulumi_gcp as gcp
net = gcp.compute.Network("net",
name="mdc-gke-net",
auto_create_subnetworks=False)
subnet = gcp.compute.Subnetwork("subnet",
name="mdc-gke-subnet",
ip_cidr_range="10.0.0.0/24",
region="us-east1",
network=net.id,
private_ip_google_access=True)
internal = gcp.compute.Firewall("internal",
name="mdc-gke-fw",
network=net.name,
direction="INGRESS",
source_ranges=["10.0.0.0/24"],
allows=[
{
"protocol": "tcp",
"ports": ["0-65535"],
},
{
"protocol": "udp",
"ports": ["0-65535"],
},
{
"protocol": "icmp",
},
])
default = gcp.vertex.AiModelDevelopmentCluster("default",
region="us-east1",
model_development_cluster_id="tftest-gke",
display_name="tftest-gke",
deletion_policy="DELETE",
labels={
"env": "dev",
},
network={
"network": net.id,
"subnetwork": subnet.id,
},
orchestrator_spec={
"kubernetes_spec": {
"node_pool_ids": ["cpu1"],
"enable_ip_alias": True,
"public_access_endpoint_enabled": True,
"control_plane_version": "-",
"node_pool_versions": {
"cpu1": "-",
},
},
},
node_pools=[{
"id": "cpu1",
"zone": "us-east1-b",
"provisioning_model": "ON_DEMAND",
"boot_disk": {
"boot_disk_size_gb": 100,
"boot_disk_type": "pd-standard",
},
"machine_spec": {
"machine_type": "n2-standard-4",
},
"scaling_spec": {
"min_node_count": 1,
"max_node_count": 2,
},
}],
opts = pulumi.ResourceOptions(depends_on=[internal]))
package main
import (
"github.com/pulumi/pulumi-gcp/sdk/v10/go/gcp/compute"
"github.com/pulumi/pulumi-gcp/sdk/v10/go/gcp/vertex"
"github.com/pulumi/pulumi/sdk/v3/go/pulumi"
)
func main() {
pulumi.Run(func(ctx *pulumi.Context) error {
net, err := compute.NewNetwork(ctx, "net", &compute.NetworkArgs{
Name: pulumi.String("mdc-gke-net"),
AutoCreateSubnetworks: pulumi.Bool(false),
})
if err != nil {
return err
}
subnet, err := compute.NewSubnetwork(ctx, "subnet", &compute.SubnetworkArgs{
Name: pulumi.String("mdc-gke-subnet"),
IpCidrRange: pulumi.String("10.0.0.0/24"),
Region: pulumi.String("us-east1"),
Network: net.ID(),
PrivateIpGoogleAccess: pulumi.Bool(true),
})
if err != nil {
return err
}
internal, err := compute.NewFirewall(ctx, "internal", &compute.FirewallArgs{
Name: pulumi.String("mdc-gke-fw"),
Network: net.Name,
Direction: pulumi.String("INGRESS"),
SourceRanges: pulumi.StringArray{
pulumi.String("10.0.0.0/24"),
},
Allows: compute.FirewallAllowArray{
&compute.FirewallAllowArgs{
Protocol: pulumi.String("tcp"),
Ports: pulumi.StringArray{
pulumi.String("0-65535"),
},
},
&compute.FirewallAllowArgs{
Protocol: pulumi.String("udp"),
Ports: pulumi.StringArray{
pulumi.String("0-65535"),
},
},
&compute.FirewallAllowArgs{
Protocol: pulumi.String("icmp"),
},
},
})
if err != nil {
return err
}
_, err = vertex.NewAiModelDevelopmentCluster(ctx, "default", &vertex.AiModelDevelopmentClusterArgs{
Region: pulumi.String("us-east1"),
ModelDevelopmentClusterId: pulumi.String("tftest-gke"),
DisplayName: pulumi.String("tftest-gke"),
DeletionPolicy: pulumi.String("DELETE"),
Labels: pulumi.StringMap{
"env": pulumi.String("dev"),
},
Network: &vertex.AiModelDevelopmentClusterNetworkArgs{
Network: net.ID(),
Subnetwork: subnet.ID(),
},
OrchestratorSpec: &vertex.AiModelDevelopmentClusterOrchestratorSpecArgs{
KubernetesSpec: &vertex.AiModelDevelopmentClusterOrchestratorSpecKubernetesSpecArgs{
NodePoolIds: pulumi.StringArray{
pulumi.String("cpu1"),
},
EnableIpAlias: pulumi.Bool(true),
PublicAccessEndpointEnabled: pulumi.Bool(true),
ControlPlaneVersion: pulumi.String("-"),
NodePoolVersions: pulumi.StringMap{
"cpu1": pulumi.String("-"),
},
},
},
NodePools: vertex.AiModelDevelopmentClusterNodePoolArray{
&vertex.AiModelDevelopmentClusterNodePoolArgs{
Id: pulumi.String("cpu1"),
Zone: pulumi.String("us-east1-b"),
ProvisioningModel: pulumi.String("ON_DEMAND"),
BootDisk: &vertex.AiModelDevelopmentClusterNodePoolBootDiskArgs{
BootDiskSizeGb: pulumi.Int(100),
BootDiskType: pulumi.String("pd-standard"),
},
MachineSpec: &vertex.AiModelDevelopmentClusterNodePoolMachineSpecArgs{
MachineType: pulumi.String("n2-standard-4"),
},
ScalingSpec: &vertex.AiModelDevelopmentClusterNodePoolScalingSpecArgs{
MinNodeCount: pulumi.Int(1),
MaxNodeCount: pulumi.Int(2),
},
},
},
}, pulumi.DependsOn([]pulumi.Resource{
internal,
}))
if err != nil {
return err
}
return nil
})
}
using System.Collections.Generic;
using System.Linq;
using Pulumi;
using Gcp = Pulumi.Gcp;
return await Deployment.RunAsync(() =>
{
var net = new Gcp.Compute.Network("net", new()
{
Name = "mdc-gke-net",
AutoCreateSubnetworks = false,
});
var subnet = new Gcp.Compute.Subnetwork("subnet", new()
{
Name = "mdc-gke-subnet",
IpCidrRange = "10.0.0.0/24",
Region = "us-east1",
Network = net.Id,
PrivateIpGoogleAccess = true,
});
var @internal = new Gcp.Compute.Firewall("internal", new()
{
Name = "mdc-gke-fw",
Network = net.Name,
Direction = "INGRESS",
SourceRanges = new[]
{
"10.0.0.0/24",
},
Allows = new[]
{
new Gcp.Compute.Inputs.FirewallAllowArgs
{
Protocol = "tcp",
Ports = new[]
{
"0-65535",
},
},
new Gcp.Compute.Inputs.FirewallAllowArgs
{
Protocol = "udp",
Ports = new[]
{
"0-65535",
},
},
new Gcp.Compute.Inputs.FirewallAllowArgs
{
Protocol = "icmp",
},
},
});
var @default = new Gcp.Vertex.AiModelDevelopmentCluster("default", new()
{
Region = "us-east1",
ModelDevelopmentClusterId = "tftest-gke",
DisplayName = "tftest-gke",
DeletionPolicy = "DELETE",
Labels =
{
{ "env", "dev" },
},
Network = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNetworkArgs
{
Network = net.Id,
Subnetwork = subnet.Id,
},
OrchestratorSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterOrchestratorSpecArgs
{
KubernetesSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterOrchestratorSpecKubernetesSpecArgs
{
NodePoolIds = new[]
{
"cpu1",
},
EnableIpAlias = true,
PublicAccessEndpointEnabled = true,
ControlPlaneVersion = "-",
NodePoolVersions =
{
{ "cpu1", "-" },
},
},
},
NodePools = new[]
{
new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolArgs
{
Id = "cpu1",
Zone = "us-east1-b",
ProvisioningModel = "ON_DEMAND",
BootDisk = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolBootDiskArgs
{
BootDiskSizeGb = 100,
BootDiskType = "pd-standard",
},
MachineSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolMachineSpecArgs
{
MachineType = "n2-standard-4",
},
ScalingSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolScalingSpecArgs
{
MinNodeCount = 1,
MaxNodeCount = 2,
},
},
},
}, new CustomResourceOptions
{
DependsOn =
{
@internal,
},
});
});
package generated_program;
import com.pulumi.Context;
import com.pulumi.Pulumi;
import com.pulumi.core.Output;
import com.pulumi.gcp.compute.Network;
import com.pulumi.gcp.compute.NetworkArgs;
import com.pulumi.gcp.compute.Subnetwork;
import com.pulumi.gcp.compute.SubnetworkArgs;
import com.pulumi.gcp.compute.Firewall;
import com.pulumi.gcp.compute.FirewallArgs;
import com.pulumi.gcp.compute.inputs.FirewallAllowArgs;
import com.pulumi.gcp.vertex.AiModelDevelopmentCluster;
import com.pulumi.gcp.vertex.AiModelDevelopmentClusterArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterNetworkArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterOrchestratorSpecArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterOrchestratorSpecKubernetesSpecArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterNodePoolArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterNodePoolBootDiskArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterNodePoolMachineSpecArgs;
import com.pulumi.gcp.vertex.inputs.AiModelDevelopmentClusterNodePoolScalingSpecArgs;
import com.pulumi.resources.CustomResourceOptions;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.Map;
import java.io.File;
import java.nio.file.Files;
import java.nio.file.Paths;
public class App {
public static void main(String[] args) {
Pulumi.run(App::stack);
}
public static void stack(Context ctx) {
var net = new Network("net", NetworkArgs.builder()
.name("mdc-gke-net")
.autoCreateSubnetworks(false)
.build());
var subnet = new Subnetwork("subnet", SubnetworkArgs.builder()
.name("mdc-gke-subnet")
.ipCidrRange("10.0.0.0/24")
.region("us-east1")
.network(net.id())
.privateIpGoogleAccess(true)
.build());
var internal = new Firewall("internal", FirewallArgs.builder()
.name("mdc-gke-fw")
.network(net.name())
.direction("INGRESS")
.sourceRanges("10.0.0.0/24")
.allows(
FirewallAllowArgs.builder()
.protocol("tcp")
.ports("0-65535")
.build(),
FirewallAllowArgs.builder()
.protocol("udp")
.ports("0-65535")
.build(),
FirewallAllowArgs.builder()
.protocol("icmp")
.build())
.build());
var default_ = new AiModelDevelopmentCluster("default", AiModelDevelopmentClusterArgs.builder()
.region("us-east1")
.modelDevelopmentClusterId("tftest-gke")
.displayName("tftest-gke")
.deletionPolicy("DELETE")
.labels(Map.of("env", "dev"))
.network(AiModelDevelopmentClusterNetworkArgs.builder()
.network(net.id())
.subnetwork(subnet.id())
.build())
.orchestratorSpec(AiModelDevelopmentClusterOrchestratorSpecArgs.builder()
.kubernetesSpec(AiModelDevelopmentClusterOrchestratorSpecKubernetesSpecArgs.builder()
.nodePoolIds("cpu1")
.enableIpAlias(true)
.publicAccessEndpointEnabled(true)
.controlPlaneVersion("-")
.nodePoolVersions(Map.of("cpu1", "-"))
.build())
.build())
.nodePools(AiModelDevelopmentClusterNodePoolArgs.builder()
.id("cpu1")
.zone("us-east1-b")
.provisioningModel("ON_DEMAND")
.bootDisk(AiModelDevelopmentClusterNodePoolBootDiskArgs.builder()
.bootDiskSizeGb(100)
.bootDiskType("pd-standard")
.build())
.machineSpec(AiModelDevelopmentClusterNodePoolMachineSpecArgs.builder()
.machineType("n2-standard-4")
.build())
.scalingSpec(AiModelDevelopmentClusterNodePoolScalingSpecArgs.builder()
.minNodeCount(1)
.maxNodeCount(2)
.build())
.build())
.build(), CustomResourceOptions.builder()
.dependsOn(internal)
.build());
}
}
resources:
net:
type: gcp:compute:Network
properties:
name: mdc-gke-net
autoCreateSubnetworks: false
subnet:
type: gcp:compute:Subnetwork
properties:
name: mdc-gke-subnet
ipCidrRange: 10.0.0.0/24
region: us-east1
network: ${net.id}
privateIpGoogleAccess: true
internal:
type: gcp:compute:Firewall
properties:
name: mdc-gke-fw
network: ${net.name}
direction: INGRESS
sourceRanges:
- 10.0.0.0/24
allows:
- protocol: tcp
ports:
- 0-65535
- protocol: udp
ports:
- 0-65535
- protocol: icmp
default:
type: gcp:vertex:AiModelDevelopmentCluster
properties:
region: us-east1
modelDevelopmentClusterId: tftest-gke
displayName: tftest-gke
deletionPolicy: DELETE
labels:
env: dev
network:
network: ${net.id}
subnetwork: ${subnet.id}
orchestratorSpec:
kubernetesSpec:
nodePoolIds:
- cpu1
enableIpAlias: true
publicAccessEndpointEnabled: true
controlPlaneVersion: '-'
nodePoolVersions:
cpu1: '-'
nodePools:
- id: cpu1
zone: us-east1-b
provisioningModel: ON_DEMAND
bootDisk:
bootDiskSizeGb: 100
bootDiskType: pd-standard
machineSpec:
machineType: n2-standard-4
scalingSpec:
minNodeCount: 1
maxNodeCount: 2
options:
dependsOn:
- ${internal}
terraform {
required_providers {
gcp = {
source = "pulumi/gcp"
}
}
}
resource "gcp_compute_network" "net" {
lifecycle {
create_before_destroy = true
}
name = "mdc-gke-net"
auto_create_subnetworks = false
}
resource "gcp_compute_subnetwork" "subnet" {
lifecycle {
create_before_destroy = true
}
name = "mdc-gke-subnet"
ip_cidr_range = "10.0.0.0/24"
region = "us-east1"
network = gcp_compute_network.net.id
private_ip_google_access = true
}
resource "gcp_compute_firewall" "internal" {
lifecycle {
create_before_destroy = true
}
name = "mdc-gke-fw"
network = gcp_compute_network.net.name
direction = "INGRESS"
source_ranges = ["10.0.0.0/24"]
allows {
protocol = "tcp"
ports = ["0-65535"]
}
allows {
protocol = "udp"
ports = ["0-65535"]
}
allows {
protocol = "icmp"
}
}
resource "gcp_vertex_ai_model_development_cluster" "default" {
depends_on = [gcp_compute_firewall.internal]
lifecycle {
create_before_destroy = true
}
region = "us-east1"
model_development_cluster_id = "tftest-gke"
display_name = "tftest-gke"
deletion_policy = "DELETE"
labels = {
"env" = "dev"
}
network = {
network = gcp_compute_network.net.id
subnetwork = gcp_compute_subnetwork.subnet.id
}
orchestrator_spec = {
kubernetes_spec = {
node_pool_ids = ["cpu1"]
enable_ip_alias = true
public_access_endpoint_enabled = true
control_plane_version = "-"
node_pool_versions = {
"cpu1" = "-"
}
}
}
node_pools {
id = "cpu1"
zone = "us-east1-b"
provisioning_model = "ON_DEMAND"
boot_disk = {
boot_disk_size_gb = 100
boot_disk_type = "pd-standard"
}
machine_spec = {
machine_type = "n2-standard-4"
}
scaling_spec = {
min_node_count = 1
max_node_count = 2
}
}
}
## Constraints:
machine_spec.reservation_affinity
must not be set if provisioningModel is ON_DEMAND, SPOT or
## FLEX_START.
machine_spec.reservation_affinity
must be set if provisioningModel is RESERVATION.
Possible values:
ON_DEMAND
RESERVATION
SPOT
FLEX_START
scalingSpec- (Optional) The scaling spec of the node pool. Structure is documented below.startupScript- (Optional) Startup script for the node pool instances.zone- (Optional) The zone of the node pool.
bootDiskSizeGb- (Optional) Size in GB of the boot disk (default is 100GB).bootDiskType- (Optional) Type of the boot disk. For non-A3U machines, the default value is “pd-ssd”, for A3U machines, the default value is “hyperdisk-balanced”. Valid values: “pd-ssd” (Persistent Disk Solid State Drive), “pd-standard” (Persistent Disk Hard Disk Drive) or “hyperdisk-balanced”.
The machineSpec block supports:
machineType- (Required) The type of the machine. See the list of machine types supported for prediction See the list of machine types supported for custom training. For DeployedModel this field is optional, and the default value isn1-standard-2. For BatchPredictionJob or as part of WorkerPoolSpec this field is required.reservationAffinity- (Optional) A ReservationAffinity can be used to configure a Vertex AI resource (e.g., a DeployedModel) to draw its Compute Engine resources from a Shared Reservation, or exclusively from on-demand capacity. Structure is documented below.
The reservationAffinity block supports:
reservationAffinityType- (Required) Specifies the reservation affinity type. Possible values: NO_RESERVATION ANY_RESERVATION SPECIFIC_RESERVATIONvalues- (Optional) Corresponds to the label values of a reservation resource. This must be the resource name of the reservation, reservation block, or reservation sub- block.
The scalingSpec block supports:
maxNodeCount- (Optional) maximum size of the node pool, must be ≥ minNodeCountminNodeCount- (Optional) minimum size of the node pool, must be >= 0 and <= maxNodeCount
The orchestratorSpec block supports:
slurmSpec- (Optional) The Slurm orchestrator related spec of the cluster. Structure is documented below.kubernetesSpec- (Optional) The GKE orchestrator related spec of the cluster. Structure is documented below.
config- (Optional) Cluster-scoped Slurm settings, emitted as globalKey=Valuelines in slurm.conf. Keys are Slurm parameter names exactly as documented by SchedMD (https://slurm.schedmd.com/slurm.conf.html), matched case-insensitively. Values are written the way slurm.conf writes them, including composite forms such asbf_window=1440,bf_resolution=600. Example: { “PriorityType”: “priority/multifactor”, “PriorityWeightAge”: “1000”, “PreemptMode”: “SUSPEND,GANG”, “SchedulerParameters”: “bf_window=1440,bf_resolution=600” } Unsupported, reserved, or misspelled parameters, and values that are not legal for the parameter, are rejected at request time with the offending parameter named; they are never silently dropped. Settings that apply to a single partition or node pool belong inSlurmPartition.configorSlurmNodeSet.configinstead - several parameters (e.g.DefMemPerCPU,OverTimeLimit,PreemptMode) are valid at more than one scope with different meanings and different legal values. On update this map is replaced wholesale; supply the complete desired set.defaultPartitionId- (Optional) Id of the partition that Slurm jobs submitted without an explicit partition run on, for examplesbatchwithout--partition. Must match theidof one ofpartitions. If unset, the service uses the first entry ofpartitionsand this field stays empty; the cluster always has a default partition whether or not this field is set. Setting it makes the choice independent of list order, which matters becausepartitionsis replaced wholesale on update: reordering the list, or prepending a partition to it, otherwise moves the default and silently redirects every job submitted without--partition.epilogBashScripts- (Optional) Custom bash scripts executed after each slurm job on every allocated node.homeDirectoryStorage- (Required) Full name of the Filestore to be mounted as /home directory on every node.loginNodePoolId- (Required) The node pool id to be used as slurm login node.nodeSets- (Optional) Node-scoped Slurm settings, keyed by node pool. At most one entry per node pool. Node pools without an entry use the settings derived by the cluster’s Slurm controller. Structure is documented below.partitions- (Optional) The Slurm partitions spec. Structure is documented below.prologBashScripts- (Optional) Custom bash scripts executed before each slurm job on every allocated node.
config- (Optional) Node-scoped Slurm settings, emitted as additionalKey=Valuepairs on theNodeName=line for this node pool’s nodes in slurm.conf. Example: { “Features”: “a3mega,nvlink”, “Weight”: “10”, “MemSpecLimit”: “4096” } On update this map is replaced wholesale; supply the complete desired set.nodePoolId- (Required) Id of the node pool these settings apply to. Must match theidof one of the node pools in the cluster.
The partitions block supports:
config- (Optional) Partition-scoped Slurm settings, emitted as additionalKey=Valuepairs on this partition’sPartitionName=line in slurm.conf. Example: { “MaxTime”: “4-00:00:00”, “DenyAccounts”: “low-pri”, “PriorityTier”: “20”, “OverSubscribe”: “YES:4” } Note: values set here take precedence over service-computed defaults for the same parameter (for exampleOverSubscribeandDefMemPerCPU, which the service otherwise derives from the partition’s exclusivity and from the memory per CPU of its nodes).Defaultis rejected here: which partition is the default is a cluster-scoped choice, so it belongs inSlurmSpec.default_partition_id.id- (Required) Id of the slurm partitonnodePoolIds- (Required) Node pools for the slurm partition
The kubernetesSpec block supports:
nodePoolIds- (Optional) IDs of the NodePools to be included in this GKE cluster. These IDs must correspond to NodePools defined in thenodePoolsfield.enableIpAlias- (Optional) Whether to enable IP alias.publicAccessEndpointEnabled- (Optional) Whether the GKE cluster’s control plane endpoint is publicly accessible. If true, the cluster is created with a public access endpoint (i.e. the private cluster configuration is disabled). If false or unset, the cluster is created with only private control plane access by default.controlPlaneVersion- (Optional) The Kubernetes version of the GKE control plane. Accepted formats are minor version (for example,"1.30"), patch version (for example,"1.30.5"), full GKE release version (for example,"1.30.5-gke.1014001"), or"-"to use the service default version. Upgrades can advance at most one minor version at a time; downgrades are rejected.nodePoolVersions- (Optional) The Kubernetes versions for specific node pools, keyed by node pool ID. When specified, all node pool IDs in the cluster must be included. Accepted formats are minor version (for example,"1.30"), patch version (for example,"1.30.5"), full GKE release version (for example,"1.30.5-gke.1014001"), or"-"to align with the control plane version. Node pools cannot be newer than the control plane or more than 2 minor versions behind it; upgrades can advance at most one minor version at a time and downgrades are rejected.
The lustreExportConfig block supports:
daysOfWeek- (Optional) Days of the week on which a backup runs, each at most once and none of them DAY_OF_WEEK_UNSPECIFIED. If empty, a backup runs every day.destinationBucket- (Required) The Cloud Storage bucket that backups are written to, without the “gs://” scheme.lustres- (Optional) The Managed Lustre instances to back up, each named as “projects/{project}/locations/{zone}/instances/{instance}”. Every entry must be attached to one of the cluster’s node pools, and no entry may repeat. If empty, every attached instance is backed up.startTime- (Optional) Represents a time of day. The date and time zone are either not significant or are specified elsewhere. An API may choose to allow leap seconds. Related types are google.type.Date andgoogle.protobuf.Timestamp. Structure is documented below.
hours- (Optional) Hours of a day in 24 hour format. Must be greater than or equal to 0 and typically must be less than or equal to 23. An API may choose to allow the value “24:00:00” for scenarios like business closing time.minutes- (Optional) Minutes of an hour. Must be greater than or equal to 0 and less than or equal to 59.
network- (Required) The name of the VPC network where to create the cluster. Thedefaultnetwork is used if not specified.subnetwork- (Optional) The name of the subnetwork in the above network to create the cluster. If not specified:- for auto mode VPC, the subnet in corresponding region is auto-selected;
- for custom mode VPC, a new subnet will be automatically created.
The runtimeSpec block supports:
serviceAccount- (Optional) The default service account used by the cluster when running any workloads on it. If not specified, the default compute engine service account will be used.
Create AiModelDevelopmentCluster Resource
Resources are created with functions called constructors. To learn more about declaring and configuring resources, see Resources.
Constructor syntax
new AiModelDevelopmentCluster(name: string, args: AiModelDevelopmentClusterArgs, opts?: CustomResourceOptions);@overload
def AiModelDevelopmentCluster(resource_name: str,
args: AiModelDevelopmentClusterArgs,
opts: Optional[ResourceOptions] = None)
@overload
def AiModelDevelopmentCluster(resource_name: str,
opts: Optional[ResourceOptions] = None,
model_development_cluster_id: Optional[str] = None,
node_pools: Optional[Sequence[AiModelDevelopmentClusterNodePoolArgs]] = None,
orchestrator_spec: Optional[AiModelDevelopmentClusterOrchestratorSpecArgs] = None,
deletion_policy: Optional[str] = None,
display_name: Optional[str] = None,
labels: Optional[Mapping[str, str]] = None,
lustre_export_config: Optional[AiModelDevelopmentClusterLustreExportConfigArgs] = None,
network: Optional[AiModelDevelopmentClusterNetworkArgs] = None,
project: Optional[str] = None,
region: Optional[str] = None,
runtime_spec: Optional[AiModelDevelopmentClusterRuntimeSpecArgs] = None,
update_mode: Optional[str] = None)func NewAiModelDevelopmentCluster(ctx *Context, name string, args AiModelDevelopmentClusterArgs, opts ...ResourceOption) (*AiModelDevelopmentCluster, error)public AiModelDevelopmentCluster(string name, AiModelDevelopmentClusterArgs args, CustomResourceOptions? opts = null)
public AiModelDevelopmentCluster(String name, AiModelDevelopmentClusterArgs args)
public AiModelDevelopmentCluster(String name, AiModelDevelopmentClusterArgs args, CustomResourceOptions options)
type: gcp:vertex:AiModelDevelopmentCluster
properties: # The arguments to resource properties.
options: # Bag of options to control resource's behavior.
resource "gcp_vertex_ai_model_development_cluster" "name" {
# resource properties
}Parameters
- name string
- The unique name of the resource.
- args AiModelDevelopmentClusterArgs
- The arguments to resource properties.
- opts CustomResourceOptions
- Bag of options to control resource's behavior.
- resource_name str
- The unique name of the resource.
- args AiModelDevelopmentClusterArgs
- The arguments to resource properties.
- opts ResourceOptions
- Bag of options to control resource's behavior.
- ctx Context
- Context object for the current deployment.
- name string
- The unique name of the resource.
- args AiModelDevelopmentClusterArgs
- The arguments to resource properties.
- opts ResourceOption
- Bag of options to control resource's behavior.
- name string
- The unique name of the resource.
- args AiModelDevelopmentClusterArgs
- The arguments to resource properties.
- opts CustomResourceOptions
- Bag of options to control resource's behavior.
- name String
- The unique name of the resource.
- args AiModelDevelopmentClusterArgs
- The arguments to resource properties.
- options CustomResourceOptions
- Bag of options to control resource's behavior.
Constructor example
The following reference example uses placeholder values for all input properties.
var aiModelDevelopmentClusterResource = new Gcp.Vertex.AiModelDevelopmentCluster("aiModelDevelopmentClusterResource", new()
{
ModelDevelopmentClusterId = "string",
NodePools = new[]
{
new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolArgs
{
Id = "string",
MachineSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolMachineSpecArgs
{
MachineType = "string",
ReservationAffinity = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolMachineSpecReservationAffinityArgs
{
ReservationAffinityType = "string",
Values = new[]
{
"string",
},
},
},
Labels =
{
{ "string", "string" },
},
Filestores = new[]
{
"string",
},
FlexStartMaxDuration = "string",
EnablePublicIps = false,
BootDisk = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolBootDiskArgs
{
BootDiskSizeGb = 0,
BootDiskType = "string",
},
Lustres = new[]
{
"string",
},
EffectiveFlexStartMaxDuration = "string",
NodeImage = "string",
ProvisioningModel = "string",
ScalingSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNodePoolScalingSpecArgs
{
MaxNodeCount = 0,
MinNodeCount = 0,
},
StartupScript = "string",
Zone = "string",
},
},
OrchestratorSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterOrchestratorSpecArgs
{
KubernetesSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterOrchestratorSpecKubernetesSpecArgs
{
ControlPlaneVersion = "string",
EnableIpAlias = false,
NodePoolIds = new[]
{
"string",
},
NodePoolVersions =
{
{ "string", "string" },
},
PublicAccessEndpointEnabled = false,
},
SlurmSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecArgs
{
HomeDirectoryStorage = "string",
LoginNodePoolId = "string",
Config =
{
{ "string", "string" },
},
DefaultPartitionId = "string",
EpilogBashScripts = new[]
{
"string",
},
NodeSets = new[]
{
new Gcp.Vertex.Inputs.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecNodeSetArgs
{
NodePoolId = "string",
Config =
{
{ "string", "string" },
},
},
},
Partitions = new[]
{
new Gcp.Vertex.Inputs.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecPartitionArgs
{
Id = "string",
NodePoolIds = new[]
{
"string",
},
Config =
{
{ "string", "string" },
},
},
},
PrologBashScripts = new[]
{
"string",
},
},
},
DeletionPolicy = "string",
DisplayName = "string",
Labels =
{
{ "string", "string" },
},
LustreExportConfig = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterLustreExportConfigArgs
{
DestinationBucket = "string",
DaysOfWeeks = new[]
{
"string",
},
Lustres = new[]
{
"string",
},
StartTime = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterLustreExportConfigStartTimeArgs
{
Hours = 0,
Minutes = 0,
},
},
Network = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterNetworkArgs
{
Network = "string",
Subnetwork = "string",
},
Project = "string",
Region = "string",
RuntimeSpec = new Gcp.Vertex.Inputs.AiModelDevelopmentClusterRuntimeSpecArgs
{
ServiceAccount = "string",
},
UpdateMode = "string",
});
example, err := vertex.NewAiModelDevelopmentCluster(ctx, "aiModelDevelopmentClusterResource", &vertex.AiModelDevelopmentClusterArgs{
ModelDevelopmentClusterId: pulumi.String("string"),
NodePools: vertex.AiModelDevelopmentClusterNodePoolArray{
&vertex.AiModelDevelopmentClusterNodePoolArgs{
Id: pulumi.String("string"),
MachineSpec: &vertex.AiModelDevelopmentClusterNodePoolMachineSpecArgs{
MachineType: pulumi.String("string"),
ReservationAffinity: &vertex.AiModelDevelopmentClusterNodePoolMachineSpecReservationAffinityArgs{
ReservationAffinityType: pulumi.String("string"),
Values: pulumi.StringArray{
pulumi.String("string"),
},
},
},
Labels: pulumi.StringMap{
"string": pulumi.String("string"),
},
Filestores: pulumi.StringArray{
pulumi.String("string"),
},
FlexStartMaxDuration: pulumi.String("string"),
EnablePublicIps: pulumi.Bool(false),
BootDisk: &vertex.AiModelDevelopmentClusterNodePoolBootDiskArgs{
BootDiskSizeGb: pulumi.Int(0),
BootDiskType: pulumi.String("string"),
},
Lustres: pulumi.StringArray{
pulumi.String("string"),
},
EffectiveFlexStartMaxDuration: pulumi.String("string"),
NodeImage: pulumi.String("string"),
ProvisioningModel: pulumi.String("string"),
ScalingSpec: &vertex.AiModelDevelopmentClusterNodePoolScalingSpecArgs{
MaxNodeCount: pulumi.Int(0),
MinNodeCount: pulumi.Int(0),
},
StartupScript: pulumi.String("string"),
Zone: pulumi.String("string"),
},
},
OrchestratorSpec: &vertex.AiModelDevelopmentClusterOrchestratorSpecArgs{
KubernetesSpec: &vertex.AiModelDevelopmentClusterOrchestratorSpecKubernetesSpecArgs{
ControlPlaneVersion: pulumi.String("string"),
EnableIpAlias: pulumi.Bool(false),
NodePoolIds: pulumi.StringArray{
pulumi.String("string"),
},
NodePoolVersions: pulumi.StringMap{
"string": pulumi.String("string"),
},
PublicAccessEndpointEnabled: pulumi.Bool(false),
},
SlurmSpec: &vertex.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecArgs{
HomeDirectoryStorage: pulumi.String("string"),
LoginNodePoolId: pulumi.String("string"),
Config: pulumi.StringMap{
"string": pulumi.String("string"),
},
DefaultPartitionId: pulumi.String("string"),
EpilogBashScripts: pulumi.StringArray{
pulumi.String("string"),
},
NodeSets: vertex.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecNodeSetArray{
&vertex.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecNodeSetArgs{
NodePoolId: pulumi.String("string"),
Config: pulumi.StringMap{
"string": pulumi.String("string"),
},
},
},
Partitions: vertex.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecPartitionArray{
&vertex.AiModelDevelopmentClusterOrchestratorSpecSlurmSpecPartitionArgs{
Id: pulumi.String("string"),
NodePoolIds: pulumi.StringArray{
pulumi.String("string"),
},
Config: pulumi.StringMap{
"string": pulumi.String("string"),
},
},
},
PrologBashScripts: pulumi.StringArray{
pulumi.String("string"),
},
},
},
DeletionPolicy: pulumi.String("string"),
DisplayName: pulumi.String("string"),
Labels: pulumi.StringMap{
"string": pulumi.String("string"),
},
LustreExportConfig: &vertex.AiModelDevelopmentClusterLustreExportConfigArgs{
DestinationBucket: pulumi.String("string"),
DaysOfWeeks: pulumi.StringArray{
pulumi.String("string"),
},
Lustres: pulumi.StringArray{
pulumi.String("string"),
},
StartTime: &vertex.AiModelDevelopmentClusterLustreExportConfigStartTimeArgs{
Hours: pulumi.Int(0),
Minutes: pulumi.Int(0),
},
},
Network: &vertex.AiModelDevelopmentClusterNetworkArgs{
Network: pulumi.String("string"),
Subnetwork: pulumi.String("string"),
},
Project: pulumi.String("string"),
Region: pulumi.String("string"),
RuntimeSpec: &vertex.AiModelDevelopmentClusterRuntimeSpecArgs{
ServiceAccount: pulumi.String("string"),
},
UpdateMode: pulumi.String("string"),
})
resource "gcp_vertex_ai_model_development_cluster" "aiModelDevelopmentClusterResource" {
lifecycle {
create_before_destroy = true
}
model_development_cluster_id = "string"
node_pools {
id = "string"
machine_spec = {
machine_type = "string"
reservation_affinity = {
reservation_affinity_type = "string"
values = ["string"]
}
}
labels = {
"string" = "string"
}
filestores = ["string"]
flex_start_max_duration = "string"
enable_public_ips = false
boot_disk = {
boot_disk_size_gb = 0
boot_disk_type = "string"
}
lustres = ["string"]
effective_flex_start_max_duration = "string"
node_image = "string"
provisioning_model = "string"
scaling_spec = {
max_node_count = 0
min_node_count = 0
}
startup_script = "string"
zone = "string"
}
orchestrator_spec = {
kubernetes_spec = {
control_plane_version = "string"
enable_ip_alias = false
node_pool_ids = ["string"]
node_pool_versions = {
"string" = "string"
}
public_access_endpoint_enabled = false
}
slurm_spec = {
home_directory_storage = "string"
login_node_pool_id = "string"
config = {
"string" = "string"
}
default_partition_id = "string"
epilog_bash_scripts = ["string"]
node_sets = [{
node_pool_id = "string"
config = {
"string" = "string"
}
}]
partitions = [{
id = "string"
node_pool_ids = ["string"]
config = {
"string" = "string"
}
}]
prolog_bash_scripts = ["string"]
}
}
deletion_policy = "string"
display_name = "string"
labels = {
"string" = "string"
}
lustre_export_config = {
destination_bucket = "string"
days_of_weeks = ["string"]
lustres = ["string"]
start_time = {
hours = 0
minutes = 0
}
}
network = {
network = "string"
subnetwork = "string"
}
project = "string"
region = "string"
runtime_spec = {
service_account = "string"
}
update_mode = "string"
}
var aiModelDevelopmentClusterResource = new AiModelDevelopmentCluster("aiModelDevelopmentClusterResource", AiModelDevelopmentClusterArgs.builder()
.modelDevelopmentClusterId("string")
.nodePools(AiModelDevelopmentClusterNodePoolArgs.builder()
.id("string")
.machineSpec(AiModelDevelopmentClusterNodePoolMachineSpecArgs.builder()
.machineType("string")
.reservationAffinity(AiModelDevelopmentClusterNodePoolMachineSpecReservationAffinityArgs.builder()
.reservationAffinityType("string")
.values("string")
.build())
.build())
.labels(Map.of("string", "string"))
.filestores("string")
.flexStartMaxDuration("string")
.enablePublicIps(false)
.bootDisk(AiModelDevelopmentClusterNodePoolBootDiskArgs.builder()
.bootDiskSizeGb(0)
.bootDiskType("string")
.build())
.lustres("string")
.effectiveFlexStartMaxDuration("string")
.nodeImage("string")
.provisioningModel("string")
.scalingSpec(AiModelDevelopmentClusterNodePoolScalingSpecArgs.builder()
.maxNodeCount(0)
.minNodeCount(0)
.build())
.startupScript("string")
.zone("string")
.build())
.orchestratorSpec(AiModelDevelopmentClusterOrchestratorSpecArgs.builder()
.kubernetesSpec(AiModelDevelopmentClusterOrchestratorSpecKubernetesSpecArgs.builder()
.controlPlaneVersion("string")
.enableIpAlias(false)
.nodePoolIds("string")
.nodePoolVersions(Map.of("string", "string"))
.publicAccessEndpointEnabled(false)
.build())
.slurmSpec(AiModelDevelopmentClusterOrchestratorSpecSlurmSpecArgs.builder()
.homeDirectoryStorage("string")
.loginNodePoolId("string")
.config(Map.of("string", "string"))
.defaultPartitionId("string")
.epilogBashScripts("string")
.nodeSets(AiModelDevelopmentClusterOrchestratorSpecSlurmSpecNodeSetArgs.builder()
.nodePoolId("string")
.config(Map.of("string", "string"))
.build())
.partitions(AiModelDevelopmentClusterOrchestratorSpecSlurmSpecPartitionArgs.builder()
.id("string")
.nodePoolIds("string")
.config(Map.of("string", "string"))
.build())
.prologBashScripts("string")
.build())
.build())
.deletionPolicy("string")
.displayName("string")
.labels(Map.of("string", "string"))
.lustreExportConfig(AiModelDevelopmentClusterLustreExportConfigArgs.builder()
.destinationBucket("string")
.daysOfWeeks("string")
.lustres("string")
.startTime(AiModelDevelopmentClusterLustreExportConfigStartTimeArgs.builder()
.hours(0)
.minutes(0)
.build())
.build())
.network(AiModelDevelopmentClusterNetworkArgs.builder()
.network("string")
.subnetwork("string")
.build())
.project("string")
.region("string")
.runtimeSpec(AiModelDevelopmentClusterRuntimeSpecArgs.builder()
.serviceAccount("string")
.build())
.updateMode("string")
.build());
ai_model_development_cluster_resource = gcp.vertex.AiModelDevelopmentCluster("aiModelDevelopmentClusterResource",
model_development_cluster_id="string",
node_pools=[{
"id": "string",
"machine_spec": {
"machine_type": "string",
"reservation_affinity": {
"reservation_affinity_type": "string",
"values": ["string"],
},
},
"labels": {
"string": "string",
},
"filestores": ["string"],
"flex_start_max_duration": "string",
"enable_public_ips": False,
"boot_disk": {
"boot_disk_size_gb": 0,
"boot_disk_type": "string",
},
"lustres": ["string"],
"effective_flex_start_max_duration": "string",
"node_image": "string",
"provisioning_model": "string",
"scaling_spec": {
"max_node_count": 0,
"min_node_count": 0,
},
"startup_script": "string",
"zone": "string",
}],
orchestrator_spec={
"kubernetes_spec": {
"control_plane_version": "string",
"enable_ip_alias": False,
"node_pool_ids": ["string"],
"node_pool_versions": {
"string": "string",
},
"public_access_endpoint_enabled": False,
},
"slurm_spec": {
"home_directory_storage": "string",
"login_node_pool_id": "string",
"config": {
"string": "string",
},
"default_partition_id": "string",
"epilog_bash_scripts": ["string"],
"node_sets": [{
"node_pool_id": "string",
"config": {
"string": "string",
},
}],
"partitions": [{
"id": "string",
"node_pool_ids": ["string"],
"config": {
"string": "string",
},
}],
"prolog_bash_scripts": ["string"],
},
},
deletion_policy="string",
display_name="string",
labels={
"string": "string",
},
lustre_export_config={
"destination_bucket": "string",
"days_of_weeks": ["string"],
"lustres": ["string"],
"start_time": {
"hours": 0,
"minutes": 0,
},
},
network={
"network": "string",
"subnetwork": "string",
},
project="string",
region="string",
runtime_spec={
"service_account": "string",
},
update_mode="string")
const aiModelDevelopmentClusterResource = new gcp.vertex.AiModelDevelopmentCluster("aiModelDevelopmentClusterResource", {
modelDevelopmentClusterId: "string",
nodePools: [{
id: "string",
machineSpec: {
machineType: "string",
reservationAffinity: {
reservationAffinityType: "string",
values: ["string"],
},
},
labels: {
string: "string",
},
filestores: ["string"],
flexStartMaxDuration: "string",
enablePublicIps: false,
bootDisk: {
bootDiskSizeGb: 0,
bootDiskType: "string",
},
lustres: ["string"],
effectiveFlexStartMaxDuration: "string",
nodeImage: "string",
provisioningModel: "string",
scalingSpec: {
maxNodeCount: 0,
minNodeCount: 0,
},
startupScript: "string",
zone: "string",
}],
orchestratorSpec: {
kubernetesSpec: {
controlPlaneVersion: "string",
enableIpAlias: false,
nodePoolIds: ["string"],
nodePoolVersions: {
string: "string",
},
publicAccessEndpointEnabled: false,
},
slurmSpec: {
homeDirectoryStorage: "string",
loginNodePoolId: "string",
config: {
string: "string",
},
defaultPartitionId: "string",
epilogBashScripts: ["string"],
nodeSets: [{
nodePoolId: "string",
config: {
string: "string",
},
}],
partitions: [{
id: "string",
nodePoolIds: ["string"],
config: {
string: "string",
},
}],
prologBashScripts: ["string"],
},
},
deletionPolicy: "string",
displayName: "string",
labels: {
string: "string",
},
lustreExportConfig: {
destinationBucket: "string",
daysOfWeeks: ["string"],
lustres: ["string"],
startTime: {
hours: 0,
minutes: 0,
},
},
network: {
network: "string",
subnetwork: "string",
},
project: "string",
region: "string",
runtimeSpec: {
serviceAccount: "string",
},
updateMode: "string",
});
type: gcp:vertex:AiModelDevelopmentCluster
properties:
deletionPolicy: string
displayName: string
labels:
string: string
lustreExportConfig:
daysOfWeeks:
- string
destinationBucket: string
lustres:
- string
startTime:
hours: 0
minutes: 0
modelDevelopmentClusterId: string
network:
network: string
subnetwork: string
nodePools:
- bootDisk:
bootDiskSizeGb: 0
bootDiskType: string
effectiveFlexStartMaxDuration: string
enablePublicIps: false
filestores:
- string
flexStartMaxDuration: string
id: string
labels:
string: string
lustres:
- string
machineSpec:
machineType: string
reservationAffinity:
reservationAffinityType: string
values:
- string
nodeImage: string
provisioningModel: string
scalingSpec:
maxNodeCount: 0
minNodeCount: 0
startupScript: string
zone: string
orchestratorSpec:
kubernetesSpec:
controlPlaneVersion: string
enableIpAlias: false
nodePoolIds:
- string
nodePoolVersions:
string: string
publicAccessEndpointEnabled: false
slurmSpec:
config:
string: string
defaultPartitionId: string
epilogBashScripts:
- string
homeDirectoryStorage: string
loginNodePoolId: string
nodeSets:
- config:
string: string
nodePoolId: string
partitions:
- config:
string: string
id: string
nodePoolIds:
- string
prologBashScripts:
- string
project: string
region: string
runtimeSpec:
serviceAccount: string
updateMode: string
AiModelDevelopmentCluster Resource Properties
To learn more about resource properties and how to use them, see Inputs and Outputs in the Architecture and Concepts docs.
Inputs
In Python, inputs that are objects can be passed either as argument classes or as dictionary literals.
The AiModelDevelopmentCluster resource accepts the following input properties:
- Model
Development stringCluster Id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - Node
Pools List<AiModel Development Cluster Node Pool> - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- Orchestrator
Spec AiModel Development Cluster Orchestrator Spec - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- Deletion
Policy string - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- Display
Name string - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- Labels Dictionary<string, string>
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - Lustre
Export AiConfig Model Development Cluster Lustre Export Config - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - Network
Ai
Model Development Cluster Network - Network spec of the cluster. Structure is documented below.
- Project string
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- Region string
- The region of the ModelDevelopmentCluster. eg us-central1
- Runtime
Spec AiModel Development Cluster Runtime Spec - The runtime spec of the cluster. Structure is documented below.
- Update
Mode string - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE.
- Model
Development stringCluster Id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - Node
Pools []AiModel Development Cluster Node Pool Args - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- Orchestrator
Spec AiModel Development Cluster Orchestrator Spec Args - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- Deletion
Policy string - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- Display
Name string - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- Labels map[string]string
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - Lustre
Export AiConfig Model Development Cluster Lustre Export Config Args - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - Network
Ai
Model Development Cluster Network Args - Network spec of the cluster. Structure is documented below.
- Project string
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- Region string
- The region of the ModelDevelopmentCluster. eg us-central1
- Runtime
Spec AiModel Development Cluster Runtime Spec Args - The runtime spec of the cluster. Structure is documented below.
- Update
Mode string - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE.
- model_
development_ stringcluster_ id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - node_
pools list(object) - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- orchestrator_
spec object - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- deletion_
policy string - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- display_
name string - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- labels map(string)
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - lustre_
export_ objectconfig - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - network object
- Network spec of the cluster. Structure is documented below.
- project string
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- region string
- The region of the ModelDevelopmentCluster. eg us-central1
- runtime_
spec object - The runtime spec of the cluster. Structure is documented below.
- update_
mode string - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE.
- model
Development StringCluster Id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - node
Pools List<AiModel Development Cluster Node Pool> - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- orchestrator
Spec AiModel Development Cluster Orchestrator Spec - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- deletion
Policy String - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- display
Name String - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- labels Map<String,String>
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - lustre
Export AiConfig Model Development Cluster Lustre Export Config - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - network
Ai
Model Development Cluster Network - Network spec of the cluster. Structure is documented below.
- project String
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- region String
- The region of the ModelDevelopmentCluster. eg us-central1
- runtime
Spec AiModel Development Cluster Runtime Spec - The runtime spec of the cluster. Structure is documented below.
- update
Mode String - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE.
- model
Development stringCluster Id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - node
Pools AiModel Development Cluster Node Pool[] - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- orchestrator
Spec AiModel Development Cluster Orchestrator Spec - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- deletion
Policy string - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- display
Name string - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- labels {[key: string]: string}
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - lustre
Export AiConfig Model Development Cluster Lustre Export Config - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - network
Ai
Model Development Cluster Network - Network spec of the cluster. Structure is documented below.
- project string
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- region string
- The region of the ModelDevelopmentCluster. eg us-central1
- runtime
Spec AiModel Development Cluster Runtime Spec - The runtime spec of the cluster. Structure is documented below.
- update
Mode string - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE.
- model_
development_ strcluster_ id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - node_
pools Sequence[AiModel Development Cluster Node Pool Args] - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- orchestrator_
spec AiModel Development Cluster Orchestrator Spec Args - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- deletion_
policy str - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- display_
name str - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- labels Mapping[str, str]
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - lustre_
export_ Aiconfig Model Development Cluster Lustre Export Config Args - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - network
Ai
Model Development Cluster Network Args - Network spec of the cluster. Structure is documented below.
- project str
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- region str
- The region of the ModelDevelopmentCluster. eg us-central1
- runtime_
spec AiModel Development Cluster Runtime Spec Args - The runtime spec of the cluster. Structure is documented below.
- update_
mode str - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE.
- model
Development StringCluster Id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - node
Pools List<Property Map> - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- orchestrator
Spec Property Map - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- deletion
Policy String - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- display
Name String - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- labels Map<String>
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - lustre
Export Property MapConfig - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - network Property Map
- Network spec of the cluster. Structure is documented below.
- project String
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- region String
- The region of the ModelDevelopmentCluster. eg us-central1
- runtime
Spec Property Map - The runtime spec of the cluster. Structure is documented below.
- update
Mode String - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE.
Outputs
All input properties are implicitly available as output properties. Additionally, the AiModelDevelopmentCluster resource produces the following output properties:
- Cluster
Events List<AiModel Development Cluster Cluster Event> - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- Cluster
Statuses List<AiModel Development Cluster Cluster Status> - The status information of the cluster. Structure is documented below.
- Create
Time string - (Output) Time when the TrainingService was created.
- Effective
Labels Dictionary<string, string> - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- Id string
- The provider-assigned unique ID for this managed resource.
- Name string
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - Pulumi
Labels Dictionary<string, string> - The combination of labels configured directly on the resource and default labels configured on the provider.
- Runtimes
List<Ai
Model Development Cluster Runtime> - The runtime backfilled records of the cluster. Structure is documented below.
- Start
Time string - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - Update
Time string - (Output) Time when the TrainingService was most recently updated.
- Cluster
Events []AiModel Development Cluster Cluster Event - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- Cluster
Statuses []AiModel Development Cluster Cluster Status - The status information of the cluster. Structure is documented below.
- Create
Time string - (Output) Time when the TrainingService was created.
- Effective
Labels map[string]string - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- Id string
- The provider-assigned unique ID for this managed resource.
- Name string
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - Pulumi
Labels map[string]string - The combination of labels configured directly on the resource and default labels configured on the provider.
- Runtimes
[]Ai
Model Development Cluster Runtime - The runtime backfilled records of the cluster. Structure is documented below.
- Start
Time string - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - Update
Time string - (Output) Time when the TrainingService was most recently updated.
- cluster_
events list(object) - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- cluster_
statuses list(object) - The status information of the cluster. Structure is documented below.
- create_
time string - (Output) Time when the TrainingService was created.
- effective_
labels map(string) - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- id string
- The provider-assigned unique ID for this managed resource.
- name string
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - pulumi_
labels map(string) - The combination of labels configured directly on the resource and default labels configured on the provider.
- runtimes list(object)
- The runtime backfilled records of the cluster. Structure is documented below.
- start_
time string - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - update_
time string - (Output) Time when the TrainingService was most recently updated.
- cluster
Events List<AiModel Development Cluster Cluster Event> - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- cluster
Statuses List<AiModel Development Cluster Cluster Status> - The status information of the cluster. Structure is documented below.
- create
Time String - (Output) Time when the TrainingService was created.
- effective
Labels Map<String,String> - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- id String
- The provider-assigned unique ID for this managed resource.
- name String
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - pulumi
Labels Map<String,String> - The combination of labels configured directly on the resource and default labels configured on the provider.
- runtimes
List<Ai
Model Development Cluster Runtime> - The runtime backfilled records of the cluster. Structure is documented below.
- start
Time String - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - update
Time String - (Output) Time when the TrainingService was most recently updated.
- cluster
Events AiModel Development Cluster Cluster Event[] - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- cluster
Statuses AiModel Development Cluster Cluster Status[] - The status information of the cluster. Structure is documented below.
- create
Time string - (Output) Time when the TrainingService was created.
- effective
Labels {[key: string]: string} - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- id string
- The provider-assigned unique ID for this managed resource.
- name string
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - pulumi
Labels {[key: string]: string} - The combination of labels configured directly on the resource and default labels configured on the provider.
- runtimes
Ai
Model Development Cluster Runtime[] - The runtime backfilled records of the cluster. Structure is documented below.
- start
Time string - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - update
Time string - (Output) Time when the TrainingService was most recently updated.
- cluster_
events Sequence[AiModel Development Cluster Cluster Event] - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- cluster_
statuses Sequence[AiModel Development Cluster Cluster Status] - The status information of the cluster. Structure is documented below.
- create_
time str - (Output) Time when the TrainingService was created.
- effective_
labels Mapping[str, str] - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- id str
- The provider-assigned unique ID for this managed resource.
- name str
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - pulumi_
labels Mapping[str, str] - The combination of labels configured directly on the resource and default labels configured on the provider.
- runtimes
Sequence[Ai
Model Development Cluster Runtime] - The runtime backfilled records of the cluster. Structure is documented below.
- start_
time str - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - update_
time str - (Output) Time when the TrainingService was most recently updated.
- cluster
Events List<Property Map> - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- cluster
Statuses List<Property Map> - The status information of the cluster. Structure is documented below.
- create
Time String - (Output) Time when the TrainingService was created.
- effective
Labels Map<String> - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- id String
- The provider-assigned unique ID for this managed resource.
- name String
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - pulumi
Labels Map<String> - The combination of labels configured directly on the resource and default labels configured on the provider.
- runtimes List<Property Map>
- The runtime backfilled records of the cluster. Structure is documented below.
- start
Time String - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - update
Time String - (Output) Time when the TrainingService was most recently updated.
Look up Existing AiModelDevelopmentCluster Resource
Get an existing AiModelDevelopmentCluster resource’s state with the given name, ID, and optional extra properties used to qualify the lookup.
public static get(name: string, id: Input<ID>, state?: AiModelDevelopmentClusterState, opts?: CustomResourceOptions): AiModelDevelopmentCluster@staticmethod
def get(resource_name: str,
id: str,
opts: Optional[ResourceOptions] = None,
cluster_events: Optional[Sequence[AiModelDevelopmentClusterClusterEventArgs]] = None,
cluster_statuses: Optional[Sequence[AiModelDevelopmentClusterClusterStatusArgs]] = None,
create_time: Optional[str] = None,
deletion_policy: Optional[str] = None,
display_name: Optional[str] = None,
effective_labels: Optional[Mapping[str, str]] = None,
labels: Optional[Mapping[str, str]] = None,
lustre_export_config: Optional[AiModelDevelopmentClusterLustreExportConfigArgs] = None,
model_development_cluster_id: Optional[str] = None,
name: Optional[str] = None,
network: Optional[AiModelDevelopmentClusterNetworkArgs] = None,
node_pools: Optional[Sequence[AiModelDevelopmentClusterNodePoolArgs]] = None,
orchestrator_spec: Optional[AiModelDevelopmentClusterOrchestratorSpecArgs] = None,
project: Optional[str] = None,
pulumi_labels: Optional[Mapping[str, str]] = None,
region: Optional[str] = None,
runtime_spec: Optional[AiModelDevelopmentClusterRuntimeSpecArgs] = None,
runtimes: Optional[Sequence[AiModelDevelopmentClusterRuntimeArgs]] = None,
start_time: Optional[str] = None,
update_mode: Optional[str] = None,
update_time: Optional[str] = None) -> AiModelDevelopmentClusterfunc GetAiModelDevelopmentCluster(ctx *Context, name string, id IDInput, state *AiModelDevelopmentClusterState, opts ...ResourceOption) (*AiModelDevelopmentCluster, error)public static AiModelDevelopmentCluster Get(string name, Input<string> id, AiModelDevelopmentClusterState? state, CustomResourceOptions? opts = null)public static AiModelDevelopmentCluster get(String name, Output<String> id, AiModelDevelopmentClusterState state, CustomResourceOptions options)resources: existingAiModelDevelopmentCluster: type: gcp:vertex:AiModelDevelopmentCluster get: id: ${id}import {
to = gcp_vertex_ai_model_development_cluster.example
id = "${id}"
}
- name string
- The unique name of the resulting resource.
- id Input<ID>
- The unique provider ID of the resource to look up.
- state AiModelDevelopmentClusterState
- Known property values the provider needs, along with the ID, to find the resource. Usually not required.
- opts CustomResourceOptions
- Bag of options to control resource's behavior.
- resource_name str
- The unique name of the resulting resource.
- id str
- The unique provider ID of the resource to look up.
- opts ResourceOptions
- Bag of options to control resource's behavior.
- ctx Context
- Context object for the current deployment.
- name string
- The unique name of the resulting resource.
- id IDInput
- The unique provider ID of the resource to look up.
- state AiModelDevelopmentClusterState
- Known property values the provider needs, along with the ID, to find the resource. Usually not required.
- opts ResourceOption
- Bag of options to control resource's behavior.
- name string
- The unique name of the resulting resource.
- id Input<string>
- The unique provider ID of the resource to look up.
- state AiModelDevelopmentClusterState
- Known property values the provider needs, along with the ID, to find the resource. Usually not required.
- opts CustomResourceOptions
- Bag of options to control resource's behavior.
- name String
- The unique name of the resulting resource.
- id Output<String>
- The unique provider ID of the resource to look up.
- state AiModelDevelopmentClusterState
- Known property values the provider needs, along with the ID, to find the resource. Usually not required.
- options CustomResourceOptions
- Bag of options to control resource's behavior.
- Cluster
Events List<AiModel Development Cluster Cluster Event> - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- Cluster
Statuses List<AiModel Development Cluster Cluster Status> - The status information of the cluster. Structure is documented below.
- Create
Time string - (Output) Time when the TrainingService was created.
- Deletion
Policy string - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- Display
Name string - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- Effective
Labels Dictionary<string, string> - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- Labels Dictionary<string, string>
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - Lustre
Export AiConfig Model Development Cluster Lustre Export Config - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - Model
Development stringCluster Id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - Name string
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - Network
Ai
Model Development Cluster Network - Network spec of the cluster. Structure is documented below.
- Node
Pools List<AiModel Development Cluster Node Pool> - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- Orchestrator
Spec AiModel Development Cluster Orchestrator Spec - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- Project string
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- Pulumi
Labels Dictionary<string, string> - The combination of labels configured directly on the resource and default labels configured on the provider.
- Region string
- The region of the ModelDevelopmentCluster. eg us-central1
- Runtime
Spec AiModel Development Cluster Runtime Spec - The runtime spec of the cluster. Structure is documented below.
- Runtimes
List<Ai
Model Development Cluster Runtime> - The runtime backfilled records of the cluster. Structure is documented below.
- Start
Time string - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - Update
Mode string - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE. - Update
Time string - (Output) Time when the TrainingService was most recently updated.
- Cluster
Events []AiModel Development Cluster Cluster Event Args - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- Cluster
Statuses []AiModel Development Cluster Cluster Status Args - The status information of the cluster. Structure is documented below.
- Create
Time string - (Output) Time when the TrainingService was created.
- Deletion
Policy string - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- Display
Name string - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- Effective
Labels map[string]string - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- Labels map[string]string
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - Lustre
Export AiConfig Model Development Cluster Lustre Export Config Args - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - Model
Development stringCluster Id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - Name string
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - Network
Ai
Model Development Cluster Network Args - Network spec of the cluster. Structure is documented below.
- Node
Pools []AiModel Development Cluster Node Pool Args - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- Orchestrator
Spec AiModel Development Cluster Orchestrator Spec Args - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- Project string
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- Pulumi
Labels map[string]string - The combination of labels configured directly on the resource and default labels configured on the provider.
- Region string
- The region of the ModelDevelopmentCluster. eg us-central1
- Runtime
Spec AiModel Development Cluster Runtime Spec Args - The runtime spec of the cluster. Structure is documented below.
- Runtimes
[]Ai
Model Development Cluster Runtime Args - The runtime backfilled records of the cluster. Structure is documented below.
- Start
Time string - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - Update
Mode string - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE. - Update
Time string - (Output) Time when the TrainingService was most recently updated.
- cluster_
events list(object) - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- cluster_
statuses list(object) - The status information of the cluster. Structure is documented below.
- create_
time string - (Output) Time when the TrainingService was created.
- deletion_
policy string - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- display_
name string - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- effective_
labels map(string) - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- labels map(string)
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - lustre_
export_ objectconfig - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - model_
development_ stringcluster_ id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - name string
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - network object
- Network spec of the cluster. Structure is documented below.
- node_
pools list(object) - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- orchestrator_
spec object - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- project string
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- pulumi_
labels map(string) - The combination of labels configured directly on the resource and default labels configured on the provider.
- region string
- The region of the ModelDevelopmentCluster. eg us-central1
- runtime_
spec object - The runtime spec of the cluster. Structure is documented below.
- runtimes list(object)
- The runtime backfilled records of the cluster. Structure is documented below.
- start_
time string - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - update_
mode string - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE. - update_
time string - (Output) Time when the TrainingService was most recently updated.
- cluster
Events List<AiModel Development Cluster Cluster Event> - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- cluster
Statuses List<AiModel Development Cluster Cluster Status> - The status information of the cluster. Structure is documented below.
- create
Time String - (Output) Time when the TrainingService was created.
- deletion
Policy String - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- display
Name String - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- effective
Labels Map<String,String> - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- labels Map<String,String>
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - lustre
Export AiConfig Model Development Cluster Lustre Export Config - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - model
Development StringCluster Id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - name String
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - network
Ai
Model Development Cluster Network - Network spec of the cluster. Structure is documented below.
- node
Pools List<AiModel Development Cluster Node Pool> - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- orchestrator
Spec AiModel Development Cluster Orchestrator Spec - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- project String
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- pulumi
Labels Map<String,String> - The combination of labels configured directly on the resource and default labels configured on the provider.
- region String
- The region of the ModelDevelopmentCluster. eg us-central1
- runtime
Spec AiModel Development Cluster Runtime Spec - The runtime spec of the cluster. Structure is documented below.
- runtimes
List<Ai
Model Development Cluster Runtime> - The runtime backfilled records of the cluster. Structure is documented below.
- start
Time String - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - update
Mode String - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE. - update
Time String - (Output) Time when the TrainingService was most recently updated.
- cluster
Events AiModel Development Cluster Cluster Event[] - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- cluster
Statuses AiModel Development Cluster Cluster Status[] - The status information of the cluster. Structure is documented below.
- create
Time string - (Output) Time when the TrainingService was created.
- deletion
Policy string - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- display
Name string - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- effective
Labels {[key: string]: string} - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- labels {[key: string]: string}
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - lustre
Export AiConfig Model Development Cluster Lustre Export Config - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - model
Development stringCluster Id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - name string
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - network
Ai
Model Development Cluster Network - Network spec of the cluster. Structure is documented below.
- node
Pools AiModel Development Cluster Node Pool[] - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- orchestrator
Spec AiModel Development Cluster Orchestrator Spec - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- project string
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- pulumi
Labels {[key: string]: string} - The combination of labels configured directly on the resource and default labels configured on the provider.
- region string
- The region of the ModelDevelopmentCluster. eg us-central1
- runtime
Spec AiModel Development Cluster Runtime Spec - The runtime spec of the cluster. Structure is documented below.
- runtimes
Ai
Model Development Cluster Runtime[] - The runtime backfilled records of the cluster. Structure is documented below.
- start
Time string - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - update
Mode string - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE. - update
Time string - (Output) Time when the TrainingService was most recently updated.
- cluster_
events Sequence[AiModel Development Cluster Cluster Event Args] - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- cluster_
statuses Sequence[AiModel Development Cluster Cluster Status Args] - The status information of the cluster. Structure is documented below.
- create_
time str - (Output) Time when the TrainingService was created.
- deletion_
policy str - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- display_
name str - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- effective_
labels Mapping[str, str] - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- labels Mapping[str, str]
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - lustre_
export_ Aiconfig Model Development Cluster Lustre Export Config Args - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - model_
development_ strcluster_ id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - name str
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - network
Ai
Model Development Cluster Network Args - Network spec of the cluster. Structure is documented below.
- node_
pools Sequence[AiModel Development Cluster Node Pool Args] - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- orchestrator_
spec AiModel Development Cluster Orchestrator Spec Args - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- project str
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- pulumi_
labels Mapping[str, str] - The combination of labels configured directly on the resource and default labels configured on the provider.
- region str
- The region of the ModelDevelopmentCluster. eg us-central1
- runtime_
spec AiModel Development Cluster Runtime Spec Args - The runtime spec of the cluster. Structure is documented below.
- runtimes
Sequence[Ai
Model Development Cluster Runtime Args] - The runtime backfilled records of the cluster. Structure is documented below.
- start_
time str - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - update_
mode str - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE. - update_
time str - (Output) Time when the TrainingService was most recently updated.
- cluster
Events List<Property Map> - The events of the cluster. Displays the 100 most recent events. Structure is documented below.
- cluster
Statuses List<Property Map> - The status information of the cluster. Structure is documented below.
- create
Time String - (Output) Time when the TrainingService was created.
- deletion
Policy String - Whether Terraform will be prevented from destroying the resource. Defaults to DELETE. When a 'terraform destroy' or 'pulumi up' would delete the resource, the command will fail if this field is set to "PREVENT" in Terraform state. When set to "ABANDON", the command will remove the resource from Terraform management without updating or deleting the resource in the API. When set to "DELETE", deleting the resource is allowed.
- display
Name String - The display name of the ModelDevelopmentCluster. The name can be up to 128 characters long and can consist of any UTF-8 characters.
- effective
Labels Map<String> - All of labels (key/value pairs) present on the resource in GCP, including the labels configured through Pulumi, other clients and services.
- labels Map<String>
- The labels with user-defined metadata to organize PersistentResource.
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters, numeric
characters, underscores and dashes. International characters are allowed.
See https://goo.gl/xmQnxf for more information and examples of labels.
Note: This field is non-authoritative, and will only manage the labels present in your configuration.
Please refer to the field
effectiveLabelsfor all of the labels present on the resource. - lustre
Export Property MapConfig - Configuration for backing up the Managed Lustre instances attached to a
cluster to a Cloud Storage bucket.
By default every instance attached to any node pool is backed up; set
lustresto narrow that to a subset. Each instance is written to its own prefix under the bucket, so several instances, and several clusters, can share a destination. Structure is documented below. - model
Development StringCluster Id - The ID to use for the cluster, which is the final component
of the ModelDevelopmentCluster's resource name.
This value should be 1-10 characters, and valid characters
are
/^a-z?$/. When usingslurmSpec, hyphens are not allowed (/^[a-z][a-z0-9]{0,9}$/). - name String
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - network Property Map
- Network spec of the cluster. Structure is documented below.
- node
Pools List<Property Map> - The spec of the pools of different resources. Node pools can be added, removed, or reordered in place; changing some fields on an existing node pool may require replacing the node pool (validated server-side by the API). Structure is documented below.
- orchestrator
Spec Property Map - The spec of the orchestrator. Only one orchestrator may be specified. Structure is documented below.
- project String
- The ID of the project in which the resource belongs. If it is not provided, the provider project is used.
- pulumi
Labels Map<String> - The combination of labels configured directly on the resource and default labels configured on the provider.
- region String
- The region of the ModelDevelopmentCluster. eg us-central1
- runtime
Spec Property Map - The runtime spec of the cluster. Structure is documented below.
- runtimes List<Property Map>
- The runtime backfilled records of the cluster. Structure is documented below.
- start
Time String - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate. - update
Mode String - Whether the service also refreshes service-managed fields (startup and
prolog/epilog scripts, default node images) on update.
USER_ONLYapplies only your changes; adding a node pool requiresUSER_AND_SERVICE. Only applies to Slurm clusters (orchestrator_spec.slurm_spec); ignored for GKE clusters (kubernetesSpec). Defaults toUSER_AND_SERVICE. Possible values areUSER_ONLYandUSER_AND_SERVICE. - update
Time String - (Output) Time when the TrainingService was most recently updated.
Supporting Types
AiModelDevelopmentClusterClusterEvent, AiModelDevelopmentClusterClusterEventArgs
- Event
Message string - (Output) The detailed description of the cluster event.
- Event
Time string - (Output) The occurrence time of the cluster event.
- Event
Type string - (Output) The type of the cluster event. Possible values: VM_PROLOG_ERROR VM_EPILOG_ERROR VM_CONTINUOUS_GPU_ERROR VM_SCAN_ERROR VM_SLURM_POWER_DOWN VM_SLURM_DRAIN VM_DELETION VM_REPORT_AND_REPLACEMENT VM_REBOOT
- Event
Message string - (Output) The detailed description of the cluster event.
- Event
Time string - (Output) The occurrence time of the cluster event.
- Event
Type string - (Output) The type of the cluster event. Possible values: VM_PROLOG_ERROR VM_EPILOG_ERROR VM_CONTINUOUS_GPU_ERROR VM_SCAN_ERROR VM_SLURM_POWER_DOWN VM_SLURM_DRAIN VM_DELETION VM_REPORT_AND_REPLACEMENT VM_REBOOT
- event_
message string - (Output) The detailed description of the cluster event.
- event_
time string - (Output) The occurrence time of the cluster event.
- event_
type string - (Output) The type of the cluster event. Possible values: VM_PROLOG_ERROR VM_EPILOG_ERROR VM_CONTINUOUS_GPU_ERROR VM_SCAN_ERROR VM_SLURM_POWER_DOWN VM_SLURM_DRAIN VM_DELETION VM_REPORT_AND_REPLACEMENT VM_REBOOT
- event
Message String - (Output) The detailed description of the cluster event.
- event
Time String - (Output) The occurrence time of the cluster event.
- event
Type String - (Output) The type of the cluster event. Possible values: VM_PROLOG_ERROR VM_EPILOG_ERROR VM_CONTINUOUS_GPU_ERROR VM_SCAN_ERROR VM_SLURM_POWER_DOWN VM_SLURM_DRAIN VM_DELETION VM_REPORT_AND_REPLACEMENT VM_REBOOT
- event
Message string - (Output) The detailed description of the cluster event.
- event
Time string - (Output) The occurrence time of the cluster event.
- event
Type string - (Output) The type of the cluster event. Possible values: VM_PROLOG_ERROR VM_EPILOG_ERROR VM_CONTINUOUS_GPU_ERROR VM_SCAN_ERROR VM_SLURM_POWER_DOWN VM_SLURM_DRAIN VM_DELETION VM_REPORT_AND_REPLACEMENT VM_REBOOT
- event_
message str - (Output) The detailed description of the cluster event.
- event_
time str - (Output) The occurrence time of the cluster event.
- event_
type str - (Output) The type of the cluster event. Possible values: VM_PROLOG_ERROR VM_EPILOG_ERROR VM_CONTINUOUS_GPU_ERROR VM_SCAN_ERROR VM_SLURM_POWER_DOWN VM_SLURM_DRAIN VM_DELETION VM_REPORT_AND_REPLACEMENT VM_REBOOT
- event
Message String - (Output) The detailed description of the cluster event.
- event
Time String - (Output) The occurrence time of the cluster event.
- event
Type String - (Output) The type of the cluster event. Possible values: VM_PROLOG_ERROR VM_EPILOG_ERROR VM_CONTINUOUS_GPU_ERROR VM_SCAN_ERROR VM_SLURM_POWER_DOWN VM_SLURM_DRAIN VM_DELETION VM_REPORT_AND_REPLACEMENT VM_REBOOT
AiModelDevelopmentClusterClusterStatus, AiModelDevelopmentClusterClusterStatusArgs
- Detail string
- (Output) Human-readable detail about the current state.
- Errors
List<Ai
Model Development Cluster Cluster Status Error> - (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - Nodes
Statuses List<AiModel Development Cluster Cluster Status Nodes Status> - (Output) The status information of all the nodes in the cluster. Structure is documented below.
- State string
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
- Detail string
- (Output) Human-readable detail about the current state.
- Errors
[]Ai
Model Development Cluster Cluster Status Error - (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - Nodes
Statuses []AiModel Development Cluster Cluster Status Nodes Status - (Output) The status information of all the nodes in the cluster. Structure is documented below.
- State string
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
- detail string
- (Output) Human-readable detail about the current state.
- errors list(object)
- (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - nodes_
statuses list(object) - (Output) The status information of all the nodes in the cluster. Structure is documented below.
- state string
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
- detail String
- (Output) Human-readable detail about the current state.
- errors
List<Ai
Model Development Cluster Cluster Status Error> - (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - nodes
Statuses List<AiModel Development Cluster Cluster Status Nodes Status> - (Output) The status information of all the nodes in the cluster. Structure is documented below.
- state String
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
- detail string
- (Output) Human-readable detail about the current state.
- errors
Ai
Model Development Cluster Cluster Status Error[] - (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - nodes
Statuses AiModel Development Cluster Cluster Status Nodes Status[] - (Output) The status information of all the nodes in the cluster. Structure is documented below.
- state string
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
- detail str
- (Output) Human-readable detail about the current state.
- errors
Sequence[Ai
Model Development Cluster Cluster Status Error] - (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - nodes_
statuses Sequence[AiModel Development Cluster Cluster Status Nodes Status] - (Output) The status information of all the nodes in the cluster. Structure is documented below.
- state str
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
- detail String
- (Output) Human-readable detail about the current state.
- errors List<Property Map>
- (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - nodes
Statuses List<Property Map> - (Output) The status information of all the nodes in the cluster. Structure is documented below.
- state String
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
AiModelDevelopmentClusterClusterStatusError, AiModelDevelopmentClusterClusterStatusErrorArgs
- Code int
- (Output) The status code, which should be an enum value of google.rpc.Code.
- Details
List<Immutable
Dictionary<string, string>> - (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- Message string
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
- Code int
- (Output) The status code, which should be an enum value of google.rpc.Code.
- Details []map[string]string
- (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- Message string
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
- code number
- (Output) The status code, which should be an enum value of google.rpc.Code.
- details list(map(string))
- (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- message string
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
- code Integer
- (Output) The status code, which should be an enum value of google.rpc.Code.
- details List<Map<String,String>>
- (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- message String
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
- code number
- (Output) The status code, which should be an enum value of google.rpc.Code.
- details {[key: string]: string}[]
- (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- message string
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
- code int
- (Output) The status code, which should be an enum value of google.rpc.Code.
- details Sequence[Mapping[str, str]]
- (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- message str
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
- code Number
- (Output) The status code, which should be an enum value of google.rpc.Code.
- details List<Map<String>>
- (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- message String
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
AiModelDevelopmentClusterClusterStatusNodesStatus, AiModelDevelopmentClusterClusterStatusNodesStatusArgs
- Create
Time string - (Output) Time when the TrainingService was created.
- Detail string
- (Output) Human-readable detail about the current state.
- Instance
Id string - (Output) The XManager instance ID (for example, "xmc-").
- Instance
Name string - (Output) The name of the Compute Engine instance associated with the node.
- Last
Update stringTime - (Output) The time when the node's current state or detail last changed.
- Node
Pool stringId - (Output) The ID of the node pool.
- Node
Pool stringType - (Output) The type of the node pool. Possible values: NODE_POOL_TYPE_LOGIN NODE_POOL_TYPE_COMPUTE
- Node
State string - (Output) The state of the node. Possible values: PREPARING RUNNING ERROR UPDATING
- Physical
Hosts List<AiModel Development Cluster Cluster Status Nodes Status Physical Host> - (Output) Physical Host Hardware information. Structure is documented below.
- Create
Time string - (Output) Time when the TrainingService was created.
- Detail string
- (Output) Human-readable detail about the current state.
- Instance
Id string - (Output) The XManager instance ID (for example, "xmc-").
- Instance
Name string - (Output) The name of the Compute Engine instance associated with the node.
- Last
Update stringTime - (Output) The time when the node's current state or detail last changed.
- Node
Pool stringId - (Output) The ID of the node pool.
- Node
Pool stringType - (Output) The type of the node pool. Possible values: NODE_POOL_TYPE_LOGIN NODE_POOL_TYPE_COMPUTE
- Node
State string - (Output) The state of the node. Possible values: PREPARING RUNNING ERROR UPDATING
- Physical
Hosts []AiModel Development Cluster Cluster Status Nodes Status Physical Host - (Output) Physical Host Hardware information. Structure is documented below.
- create_
time string - (Output) Time when the TrainingService was created.
- detail string
- (Output) Human-readable detail about the current state.
- instance_
id string - (Output) The XManager instance ID (for example, "xmc-").
- instance_
name string - (Output) The name of the Compute Engine instance associated with the node.
- last_
update_ stringtime - (Output) The time when the node's current state or detail last changed.
- node_
pool_ stringid - (Output) The ID of the node pool.
- node_
pool_ stringtype - (Output) The type of the node pool. Possible values: NODE_POOL_TYPE_LOGIN NODE_POOL_TYPE_COMPUTE
- node_
state string - (Output) The state of the node. Possible values: PREPARING RUNNING ERROR UPDATING
- physical_
hosts list(object) - (Output) Physical Host Hardware information. Structure is documented below.
- create
Time String - (Output) Time when the TrainingService was created.
- detail String
- (Output) Human-readable detail about the current state.
- instance
Id String - (Output) The XManager instance ID (for example, "xmc-").
- instance
Name String - (Output) The name of the Compute Engine instance associated with the node.
- last
Update StringTime - (Output) The time when the node's current state or detail last changed.
- node
Pool StringId - (Output) The ID of the node pool.
- node
Pool StringType - (Output) The type of the node pool. Possible values: NODE_POOL_TYPE_LOGIN NODE_POOL_TYPE_COMPUTE
- node
State String - (Output) The state of the node. Possible values: PREPARING RUNNING ERROR UPDATING
- physical
Hosts List<AiModel Development Cluster Cluster Status Nodes Status Physical Host> - (Output) Physical Host Hardware information. Structure is documented below.
- create
Time string - (Output) Time when the TrainingService was created.
- detail string
- (Output) Human-readable detail about the current state.
- instance
Id string - (Output) The XManager instance ID (for example, "xmc-").
- instance
Name string - (Output) The name of the Compute Engine instance associated with the node.
- last
Update stringTime - (Output) The time when the node's current state or detail last changed.
- node
Pool stringId - (Output) The ID of the node pool.
- node
Pool stringType - (Output) The type of the node pool. Possible values: NODE_POOL_TYPE_LOGIN NODE_POOL_TYPE_COMPUTE
- node
State string - (Output) The state of the node. Possible values: PREPARING RUNNING ERROR UPDATING
- physical
Hosts AiModel Development Cluster Cluster Status Nodes Status Physical Host[] - (Output) Physical Host Hardware information. Structure is documented below.
- create_
time str - (Output) Time when the TrainingService was created.
- detail str
- (Output) Human-readable detail about the current state.
- instance_
id str - (Output) The XManager instance ID (for example, "xmc-").
- instance_
name str - (Output) The name of the Compute Engine instance associated with the node.
- last_
update_ strtime - (Output) The time when the node's current state or detail last changed.
- node_
pool_ strid - (Output) The ID of the node pool.
- node_
pool_ strtype - (Output) The type of the node pool. Possible values: NODE_POOL_TYPE_LOGIN NODE_POOL_TYPE_COMPUTE
- node_
state str - (Output) The state of the node. Possible values: PREPARING RUNNING ERROR UPDATING
- physical_
hosts Sequence[AiModel Development Cluster Cluster Status Nodes Status Physical Host] - (Output) Physical Host Hardware information. Structure is documented below.
- create
Time String - (Output) Time when the TrainingService was created.
- detail String
- (Output) Human-readable detail about the current state.
- instance
Id String - (Output) The XManager instance ID (for example, "xmc-").
- instance
Name String - (Output) The name of the Compute Engine instance associated with the node.
- last
Update StringTime - (Output) The time when the node's current state or detail last changed.
- node
Pool StringId - (Output) The ID of the node pool.
- node
Pool StringType - (Output) The type of the node pool. Possible values: NODE_POOL_TYPE_LOGIN NODE_POOL_TYPE_COMPUTE
- node
State String - (Output) The state of the node. Possible values: PREPARING RUNNING ERROR UPDATING
- physical
Hosts List<Property Map> - (Output) Physical Host Hardware information. Structure is documented below.
AiModelDevelopmentClusterClusterStatusNodesStatusPhysicalHost, AiModelDevelopmentClusterClusterStatusNodesStatusPhysicalHostArgs
- Block
Id string - (Output) Name of the block.
- Cluster
Name string - (Output) Name of the physical cluster.
- Host
Id string - (Output) Name of the host.
- Subblock
Id string - (Output) Name of the sub-block.
- Block
Id string - (Output) Name of the block.
- Cluster
Name string - (Output) Name of the physical cluster.
- Host
Id string - (Output) Name of the host.
- Subblock
Id string - (Output) Name of the sub-block.
- block_
id string - (Output) Name of the block.
- cluster_
name string - (Output) Name of the physical cluster.
- host_
id string - (Output) Name of the host.
- subblock_
id string - (Output) Name of the sub-block.
- block
Id String - (Output) Name of the block.
- cluster
Name String - (Output) Name of the physical cluster.
- host
Id String - (Output) Name of the host.
- subblock
Id String - (Output) Name of the sub-block.
- block
Id string - (Output) Name of the block.
- cluster
Name string - (Output) Name of the physical cluster.
- host
Id string - (Output) Name of the host.
- subblock
Id string - (Output) Name of the sub-block.
- block_
id str - (Output) Name of the block.
- cluster_
name str - (Output) Name of the physical cluster.
- host_
id str - (Output) Name of the host.
- subblock_
id str - (Output) Name of the sub-block.
- block
Id String - (Output) Name of the block.
- cluster
Name String - (Output) Name of the physical cluster.
- host
Id String - (Output) Name of the host.
- subblock
Id String - (Output) Name of the sub-block.
AiModelDevelopmentClusterLustreExportConfig, AiModelDevelopmentClusterLustreExportConfigArgs
- Destination
Bucket string - The Cloud Storage bucket that backups are written to, without the "gs://" scheme.
- Days
Of List<string>Weeks - Days of the week on which a backup runs, each at most once and none of them DAY_OF_WEEK_UNSPECIFIED. If empty, a backup runs every day.
- Lustres List<string>
- The Managed Lustre instances to back up, each named as "projects/{project}/locations/{zone}/instances/{instance}". Every entry must be attached to one of the cluster's node pools, and no entry may repeat. If empty, every attached instance is backed up.
- Start
Time AiModel Development Cluster Lustre Export Config Start Time - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate.
- Destination
Bucket string - The Cloud Storage bucket that backups are written to, without the "gs://" scheme.
- Days
Of []stringWeeks - Days of the week on which a backup runs, each at most once and none of them DAY_OF_WEEK_UNSPECIFIED. If empty, a backup runs every day.
- Lustres []string
- The Managed Lustre instances to back up, each named as "projects/{project}/locations/{zone}/instances/{instance}". Every entry must be attached to one of the cluster's node pools, and no entry may repeat. If empty, every attached instance is backed up.
- Start
Time AiModel Development Cluster Lustre Export Config Start Time - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate.
- destination_
bucket string - The Cloud Storage bucket that backups are written to, without the "gs://" scheme.
- days_
of_ list(string)weeks - Days of the week on which a backup runs, each at most once and none of them DAY_OF_WEEK_UNSPECIFIED. If empty, a backup runs every day.
- lustres list(string)
- The Managed Lustre instances to back up, each named as "projects/{project}/locations/{zone}/instances/{instance}". Every entry must be attached to one of the cluster's node pools, and no entry may repeat. If empty, every attached instance is backed up.
- start_
time object - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate.
- destination
Bucket String - The Cloud Storage bucket that backups are written to, without the "gs://" scheme.
- days
Of List<String>Weeks - Days of the week on which a backup runs, each at most once and none of them DAY_OF_WEEK_UNSPECIFIED. If empty, a backup runs every day.
- lustres List<String>
- The Managed Lustre instances to back up, each named as "projects/{project}/locations/{zone}/instances/{instance}". Every entry must be attached to one of the cluster's node pools, and no entry may repeat. If empty, every attached instance is backed up.
- start
Time AiModel Development Cluster Lustre Export Config Start Time - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate.
- destination
Bucket string - The Cloud Storage bucket that backups are written to, without the "gs://" scheme.
- days
Of string[]Weeks - Days of the week on which a backup runs, each at most once and none of them DAY_OF_WEEK_UNSPECIFIED. If empty, a backup runs every day.
- lustres string[]
- The Managed Lustre instances to back up, each named as "projects/{project}/locations/{zone}/instances/{instance}". Every entry must be attached to one of the cluster's node pools, and no entry may repeat. If empty, every attached instance is backed up.
- start
Time AiModel Development Cluster Lustre Export Config Start Time - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate.
- destination_
bucket str - The Cloud Storage bucket that backups are written to, without the "gs://" scheme.
- days_
of_ Sequence[str]weeks - Days of the week on which a backup runs, each at most once and none of them DAY_OF_WEEK_UNSPECIFIED. If empty, a backup runs every day.
- lustres Sequence[str]
- The Managed Lustre instances to back up, each named as "projects/{project}/locations/{zone}/instances/{instance}". Every entry must be attached to one of the cluster's node pools, and no entry may repeat. If empty, every attached instance is backed up.
- start_
time AiModel Development Cluster Lustre Export Config Start Time - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate.
- destination
Bucket String - The Cloud Storage bucket that backups are written to, without the "gs://" scheme.
- days
Of List<String>Weeks - Days of the week on which a backup runs, each at most once and none of them DAY_OF_WEEK_UNSPECIFIED. If empty, a backup runs every day.
- lustres List<String>
- The Managed Lustre instances to back up, each named as "projects/{project}/locations/{zone}/instances/{instance}". Every entry must be attached to one of the cluster's node pools, and no entry may repeat. If empty, every attached instance is backed up.
- start
Time Property Map - Time when the ModelDevelopmentCluster for the first time entered the
RUNNINGstate.
AiModelDevelopmentClusterLustreExportConfigStartTime, AiModelDevelopmentClusterLustreExportConfigStartTimeArgs
- Hours int
- Hours of a day in 24 hour format. Must be greater than or equal to 0 and typically must be less than or equal to 23. An API may choose to allow the value "24:00:00" for scenarios like business closing time.
- Minutes int
- Minutes of an hour. Must be greater than or equal to 0 and less than or equal to 59.
- Hours int
- Hours of a day in 24 hour format. Must be greater than or equal to 0 and typically must be less than or equal to 23. An API may choose to allow the value "24:00:00" for scenarios like business closing time.
- Minutes int
- Minutes of an hour. Must be greater than or equal to 0 and less than or equal to 59.
- hours number
- Hours of a day in 24 hour format. Must be greater than or equal to 0 and typically must be less than or equal to 23. An API may choose to allow the value "24:00:00" for scenarios like business closing time.
- minutes number
- Minutes of an hour. Must be greater than or equal to 0 and less than or equal to 59.
- hours Integer
- Hours of a day in 24 hour format. Must be greater than or equal to 0 and typically must be less than or equal to 23. An API may choose to allow the value "24:00:00" for scenarios like business closing time.
- minutes Integer
- Minutes of an hour. Must be greater than or equal to 0 and less than or equal to 59.
- hours number
- Hours of a day in 24 hour format. Must be greater than or equal to 0 and typically must be less than or equal to 23. An API may choose to allow the value "24:00:00" for scenarios like business closing time.
- minutes number
- Minutes of an hour. Must be greater than or equal to 0 and less than or equal to 59.
- hours int
- Hours of a day in 24 hour format. Must be greater than or equal to 0 and typically must be less than or equal to 23. An API may choose to allow the value "24:00:00" for scenarios like business closing time.
- minutes int
- Minutes of an hour. Must be greater than or equal to 0 and less than or equal to 59.
- hours Number
- Hours of a day in 24 hour format. Must be greater than or equal to 0 and typically must be less than or equal to 23. An API may choose to allow the value "24:00:00" for scenarios like business closing time.
- minutes Number
- Minutes of an hour. Must be greater than or equal to 0 and less than or equal to 59.
AiModelDevelopmentClusterNetwork, AiModelDevelopmentClusterNetworkArgs
- Network string
- Network spec of the cluster. Structure is documented below.
- Subnetwork string
- The name of the subnetwork in the above network to create the cluster.
If not specified:
- for auto mode VPC, the subnet in corresponding region is auto-selected;
- for custom mode VPC, a new subnet will be automatically created.
- Network string
- Network spec of the cluster. Structure is documented below.
- Subnetwork string
- The name of the subnetwork in the above network to create the cluster.
If not specified:
- for auto mode VPC, the subnet in corresponding region is auto-selected;
- for custom mode VPC, a new subnet will be automatically created.
- network string
- Network spec of the cluster. Structure is documented below.
- subnetwork string
- The name of the subnetwork in the above network to create the cluster.
If not specified:
- for auto mode VPC, the subnet in corresponding region is auto-selected;
- for custom mode VPC, a new subnet will be automatically created.
- network String
- Network spec of the cluster. Structure is documented below.
- subnetwork String
- The name of the subnetwork in the above network to create the cluster.
If not specified:
- for auto mode VPC, the subnet in corresponding region is auto-selected;
- for custom mode VPC, a new subnet will be automatically created.
- network string
- Network spec of the cluster. Structure is documented below.
- subnetwork string
- The name of the subnetwork in the above network to create the cluster.
If not specified:
- for auto mode VPC, the subnet in corresponding region is auto-selected;
- for custom mode VPC, a new subnet will be automatically created.
- network str
- Network spec of the cluster. Structure is documented below.
- subnetwork str
- The name of the subnetwork in the above network to create the cluster.
If not specified:
- for auto mode VPC, the subnet in corresponding region is auto-selected;
- for custom mode VPC, a new subnet will be automatically created.
- network String
- Network spec of the cluster. Structure is documented below.
- subnetwork String
- The name of the subnetwork in the above network to create the cluster.
If not specified:
- for auto mode VPC, the subnet in corresponding region is auto-selected;
- for custom mode VPC, a new subnet will be automatically created.
AiModelDevelopmentClusterNodePool, AiModelDevelopmentClusterNodePoolArgs
- Id string
- The unique ID referring to this node pool. Cannot be modified in
place on an existing node pool; changing
idupdates the list of node pools by adding/removing pools (validated server-side). - Machine
Spec AiModel Development Cluster Node Pool Machine Spec - Specification of a single machine. Structure is documented below.
- Boot
Disk AiModel Development Cluster Node Pool Boot Disk - Represents the spec of disk options. Structure is documented below.
- Effective
Flex stringStart Max Duration - (Output)
The effective maximum duration for instances in this node pool.
This field reflects the value of
flexStartMaxDurationif it is set by the user, or the server-applied default (for example, 7 days) if it is not. This field is only applicable whenprovisioningModelis set toFLEX_START. - Enable
Public boolIps - Whether to enable public IPs for the node pool.
- Filestores List<string>
- List of the existing Filestore instances to be mounted. For example: if "projects/my-project/locations/us-central1-a/instances/my-instance" is specified, the instance will be mounted to the node pool under the path "mnt/filestore/"
- Flex
Start stringMax Duration - The maximum duration for instances in this node pool.
This field is only applicable when
provisioningModelis set toFLEX_START. - Labels Dictionary<string, string>
- The labels with user-defined metadata to organize the node pool. These labels are applied to the underlying compute resources (for example, the VM instances) of the node pool, so they appear in billing exports and Cloud Monitoring filters. Label keys and values can be no longer than 64 characters (Unicode codepoints), can only contain lowercase letters, numeric characters, underscores and dashes. International characters are allowed. See https://goo.gl/xmQnxf for more information and examples of labels.
- Lustres List<string>
- List of the existing Google Managed Lustre instances to be mounted. Similar to the filestores, the instance will be mounted to the node pool under the path "/mnt/lustre/"
- Node
Image string - The node image to be used for the VMs in the node pool. In format of "projects//global/images/" or "projects/test-project/global/images/family/"
- Provisioning
Model string - Specifies the provisioning model for the node pool.
Behavior for
PROVISIONING_MODEL_UNSPECIFIED:- For CPU node pools: Defaults to
ON_DEMAND. - For GPU node pools:
- If
machine_spec.reservation_affinityis set, it defaults toRESERVATION. - If
machine_spec.reservation_affinityis NOT set, this is an INVALID state and the request will be rejected. An explicitprovisioningModelmust be provided.
- For CPU node pools: Defaults to
- Scaling
Spec AiModel Development Cluster Node Pool Scaling Spec - The scaling spec of the node pool.
- Startup
Script string - Startup script for the node pool instances.
- Zone string
- The zone of the node pool.
- Id string
- The unique ID referring to this node pool. Cannot be modified in
place on an existing node pool; changing
idupdates the list of node pools by adding/removing pools (validated server-side). - Machine
Spec AiModel Development Cluster Node Pool Machine Spec - Specification of a single machine. Structure is documented below.
- Boot
Disk AiModel Development Cluster Node Pool Boot Disk - Represents the spec of disk options. Structure is documented below.
- Effective
Flex stringStart Max Duration - (Output)
The effective maximum duration for instances in this node pool.
This field reflects the value of
flexStartMaxDurationif it is set by the user, or the server-applied default (for example, 7 days) if it is not. This field is only applicable whenprovisioningModelis set toFLEX_START. - Enable
Public boolIps - Whether to enable public IPs for the node pool.
- Filestores []string
- List of the existing Filestore instances to be mounted. For example: if "projects/my-project/locations/us-central1-a/instances/my-instance" is specified, the instance will be mounted to the node pool under the path "mnt/filestore/"
- Flex
Start stringMax Duration - The maximum duration for instances in this node pool.
This field is only applicable when
provisioningModelis set toFLEX_START. - Labels map[string]string
- The labels with user-defined metadata to organize the node pool. These labels are applied to the underlying compute resources (for example, the VM instances) of the node pool, so they appear in billing exports and Cloud Monitoring filters. Label keys and values can be no longer than 64 characters (Unicode codepoints), can only contain lowercase letters, numeric characters, underscores and dashes. International characters are allowed. See https://goo.gl/xmQnxf for more information and examples of labels.
- Lustres []string
- List of the existing Google Managed Lustre instances to be mounted. Similar to the filestores, the instance will be mounted to the node pool under the path "/mnt/lustre/"
- Node
Image string - The node image to be used for the VMs in the node pool. In format of "projects//global/images/" or "projects/test-project/global/images/family/"
- Provisioning
Model string - Specifies the provisioning model for the node pool.
Behavior for
PROVISIONING_MODEL_UNSPECIFIED:- For CPU node pools: Defaults to
ON_DEMAND. - For GPU node pools:
- If
machine_spec.reservation_affinityis set, it defaults toRESERVATION. - If
machine_spec.reservation_affinityis NOT set, this is an INVALID state and the request will be rejected. An explicitprovisioningModelmust be provided.
- For CPU node pools: Defaults to
- Scaling
Spec AiModel Development Cluster Node Pool Scaling Spec - The scaling spec of the node pool.
- Startup
Script string - Startup script for the node pool instances.
- Zone string
- The zone of the node pool.
- id string
- The unique ID referring to this node pool. Cannot be modified in
place on an existing node pool; changing
idupdates the list of node pools by adding/removing pools (validated server-side). - machine_
spec object - Specification of a single machine. Structure is documented below.
- boot_
disk object - Represents the spec of disk options. Structure is documented below.
- effective_
flex_ stringstart_ max_ duration - (Output)
The effective maximum duration for instances in this node pool.
This field reflects the value of
flexStartMaxDurationif it is set by the user, or the server-applied default (for example, 7 days) if it is not. This field is only applicable whenprovisioningModelis set toFLEX_START. - enable_
public_ boolips - Whether to enable public IPs for the node pool.
- filestores list(string)
- List of the existing Filestore instances to be mounted. For example: if "projects/my-project/locations/us-central1-a/instances/my-instance" is specified, the instance will be mounted to the node pool under the path "mnt/filestore/"
- flex_
start_ stringmax_ duration - The maximum duration for instances in this node pool.
This field is only applicable when
provisioningModelis set toFLEX_START. - labels map(string)
- The labels with user-defined metadata to organize the node pool. These labels are applied to the underlying compute resources (for example, the VM instances) of the node pool, so they appear in billing exports and Cloud Monitoring filters. Label keys and values can be no longer than 64 characters (Unicode codepoints), can only contain lowercase letters, numeric characters, underscores and dashes. International characters are allowed. See https://goo.gl/xmQnxf for more information and examples of labels.
- lustres list(string)
- List of the existing Google Managed Lustre instances to be mounted. Similar to the filestores, the instance will be mounted to the node pool under the path "/mnt/lustre/"
- node_
image string - The node image to be used for the VMs in the node pool. In format of "projects//global/images/" or "projects/test-project/global/images/family/"
- provisioning_
model string - Specifies the provisioning model for the node pool.
Behavior for
PROVISIONING_MODEL_UNSPECIFIED:- For CPU node pools: Defaults to
ON_DEMAND. - For GPU node pools:
- If
machine_spec.reservation_affinityis set, it defaults toRESERVATION. - If
machine_spec.reservation_affinityis NOT set, this is an INVALID state and the request will be rejected. An explicitprovisioningModelmust be provided.
- For CPU node pools: Defaults to
- scaling_
spec object - The scaling spec of the node pool.
- startup_
script string - Startup script for the node pool instances.
- zone string
- The zone of the node pool.
- id String
- The unique ID referring to this node pool. Cannot be modified in
place on an existing node pool; changing
idupdates the list of node pools by adding/removing pools (validated server-side). - machine
Spec AiModel Development Cluster Node Pool Machine Spec - Specification of a single machine. Structure is documented below.
- boot
Disk AiModel Development Cluster Node Pool Boot Disk - Represents the spec of disk options. Structure is documented below.
- effective
Flex StringStart Max Duration - (Output)
The effective maximum duration for instances in this node pool.
This field reflects the value of
flexStartMaxDurationif it is set by the user, or the server-applied default (for example, 7 days) if it is not. This field is only applicable whenprovisioningModelis set toFLEX_START. - enable
Public BooleanIps - Whether to enable public IPs for the node pool.
- filestores List<String>
- List of the existing Filestore instances to be mounted. For example: if "projects/my-project/locations/us-central1-a/instances/my-instance" is specified, the instance will be mounted to the node pool under the path "mnt/filestore/"
- flex
Start StringMax Duration - The maximum duration for instances in this node pool.
This field is only applicable when
provisioningModelis set toFLEX_START. - labels Map<String,String>
- The labels with user-defined metadata to organize the node pool. These labels are applied to the underlying compute resources (for example, the VM instances) of the node pool, so they appear in billing exports and Cloud Monitoring filters. Label keys and values can be no longer than 64 characters (Unicode codepoints), can only contain lowercase letters, numeric characters, underscores and dashes. International characters are allowed. See https://goo.gl/xmQnxf for more information and examples of labels.
- lustres List<String>
- List of the existing Google Managed Lustre instances to be mounted. Similar to the filestores, the instance will be mounted to the node pool under the path "/mnt/lustre/"
- node
Image String - The node image to be used for the VMs in the node pool. In format of "projects//global/images/" or "projects/test-project/global/images/family/"
- provisioning
Model String - Specifies the provisioning model for the node pool.
Behavior for
PROVISIONING_MODEL_UNSPECIFIED:- For CPU node pools: Defaults to
ON_DEMAND. - For GPU node pools:
- If
machine_spec.reservation_affinityis set, it defaults toRESERVATION. - If
machine_spec.reservation_affinityis NOT set, this is an INVALID state and the request will be rejected. An explicitprovisioningModelmust be provided.
- For CPU node pools: Defaults to
- scaling
Spec AiModel Development Cluster Node Pool Scaling Spec - The scaling spec of the node pool.
- startup
Script String - Startup script for the node pool instances.
- zone String
- The zone of the node pool.
- id string
- The unique ID referring to this node pool. Cannot be modified in
place on an existing node pool; changing
idupdates the list of node pools by adding/removing pools (validated server-side). - machine
Spec AiModel Development Cluster Node Pool Machine Spec - Specification of a single machine. Structure is documented below.
- boot
Disk AiModel Development Cluster Node Pool Boot Disk - Represents the spec of disk options. Structure is documented below.
- effective
Flex stringStart Max Duration - (Output)
The effective maximum duration for instances in this node pool.
This field reflects the value of
flexStartMaxDurationif it is set by the user, or the server-applied default (for example, 7 days) if it is not. This field is only applicable whenprovisioningModelis set toFLEX_START. - enable
Public booleanIps - Whether to enable public IPs for the node pool.
- filestores string[]
- List of the existing Filestore instances to be mounted. For example: if "projects/my-project/locations/us-central1-a/instances/my-instance" is specified, the instance will be mounted to the node pool under the path "mnt/filestore/"
- flex
Start stringMax Duration - The maximum duration for instances in this node pool.
This field is only applicable when
provisioningModelis set toFLEX_START. - labels {[key: string]: string}
- The labels with user-defined metadata to organize the node pool. These labels are applied to the underlying compute resources (for example, the VM instances) of the node pool, so they appear in billing exports and Cloud Monitoring filters. Label keys and values can be no longer than 64 characters (Unicode codepoints), can only contain lowercase letters, numeric characters, underscores and dashes. International characters are allowed. See https://goo.gl/xmQnxf for more information and examples of labels.
- lustres string[]
- List of the existing Google Managed Lustre instances to be mounted. Similar to the filestores, the instance will be mounted to the node pool under the path "/mnt/lustre/"
- node
Image string - The node image to be used for the VMs in the node pool. In format of "projects//global/images/" or "projects/test-project/global/images/family/"
- provisioning
Model string - Specifies the provisioning model for the node pool.
Behavior for
PROVISIONING_MODEL_UNSPECIFIED:- For CPU node pools: Defaults to
ON_DEMAND. - For GPU node pools:
- If
machine_spec.reservation_affinityis set, it defaults toRESERVATION. - If
machine_spec.reservation_affinityis NOT set, this is an INVALID state and the request will be rejected. An explicitprovisioningModelmust be provided.
- For CPU node pools: Defaults to
- scaling
Spec AiModel Development Cluster Node Pool Scaling Spec - The scaling spec of the node pool.
- startup
Script string - Startup script for the node pool instances.
- zone string
- The zone of the node pool.
- id str
- The unique ID referring to this node pool. Cannot be modified in
place on an existing node pool; changing
idupdates the list of node pools by adding/removing pools (validated server-side). - machine_
spec AiModel Development Cluster Node Pool Machine Spec - Specification of a single machine. Structure is documented below.
- boot_
disk AiModel Development Cluster Node Pool Boot Disk - Represents the spec of disk options. Structure is documented below.
- effective_
flex_ strstart_ max_ duration - (Output)
The effective maximum duration for instances in this node pool.
This field reflects the value of
flexStartMaxDurationif it is set by the user, or the server-applied default (for example, 7 days) if it is not. This field is only applicable whenprovisioningModelis set toFLEX_START. - enable_
public_ boolips - Whether to enable public IPs for the node pool.
- filestores Sequence[str]
- List of the existing Filestore instances to be mounted. For example: if "projects/my-project/locations/us-central1-a/instances/my-instance" is specified, the instance will be mounted to the node pool under the path "mnt/filestore/"
- flex_
start_ strmax_ duration - The maximum duration for instances in this node pool.
This field is only applicable when
provisioningModelis set toFLEX_START. - labels Mapping[str, str]
- The labels with user-defined metadata to organize the node pool. These labels are applied to the underlying compute resources (for example, the VM instances) of the node pool, so they appear in billing exports and Cloud Monitoring filters. Label keys and values can be no longer than 64 characters (Unicode codepoints), can only contain lowercase letters, numeric characters, underscores and dashes. International characters are allowed. See https://goo.gl/xmQnxf for more information and examples of labels.
- lustres Sequence[str]
- List of the existing Google Managed Lustre instances to be mounted. Similar to the filestores, the instance will be mounted to the node pool under the path "/mnt/lustre/"
- node_
image str - The node image to be used for the VMs in the node pool. In format of "projects//global/images/" or "projects/test-project/global/images/family/"
- provisioning_
model str - Specifies the provisioning model for the node pool.
Behavior for
PROVISIONING_MODEL_UNSPECIFIED:- For CPU node pools: Defaults to
ON_DEMAND. - For GPU node pools:
- If
machine_spec.reservation_affinityis set, it defaults toRESERVATION. - If
machine_spec.reservation_affinityis NOT set, this is an INVALID state and the request will be rejected. An explicitprovisioningModelmust be provided.
- For CPU node pools: Defaults to
- scaling_
spec AiModel Development Cluster Node Pool Scaling Spec - The scaling spec of the node pool.
- startup_
script str - Startup script for the node pool instances.
- zone str
- The zone of the node pool.
- id String
- The unique ID referring to this node pool. Cannot be modified in
place on an existing node pool; changing
idupdates the list of node pools by adding/removing pools (validated server-side). - machine
Spec Property Map - Specification of a single machine. Structure is documented below.
- boot
Disk Property Map - Represents the spec of disk options. Structure is documented below.
- effective
Flex StringStart Max Duration - (Output)
The effective maximum duration for instances in this node pool.
This field reflects the value of
flexStartMaxDurationif it is set by the user, or the server-applied default (for example, 7 days) if it is not. This field is only applicable whenprovisioningModelis set toFLEX_START. - enable
Public BooleanIps - Whether to enable public IPs for the node pool.
- filestores List<String>
- List of the existing Filestore instances to be mounted. For example: if "projects/my-project/locations/us-central1-a/instances/my-instance" is specified, the instance will be mounted to the node pool under the path "mnt/filestore/"
- flex
Start StringMax Duration - The maximum duration for instances in this node pool.
This field is only applicable when
provisioningModelis set toFLEX_START. - labels Map<String>
- The labels with user-defined metadata to organize the node pool. These labels are applied to the underlying compute resources (for example, the VM instances) of the node pool, so they appear in billing exports and Cloud Monitoring filters. Label keys and values can be no longer than 64 characters (Unicode codepoints), can only contain lowercase letters, numeric characters, underscores and dashes. International characters are allowed. See https://goo.gl/xmQnxf for more information and examples of labels.
- lustres List<String>
- List of the existing Google Managed Lustre instances to be mounted. Similar to the filestores, the instance will be mounted to the node pool under the path "/mnt/lustre/"
- node
Image String - The node image to be used for the VMs in the node pool. In format of "projects//global/images/" or "projects/test-project/global/images/family/"
- provisioning
Model String - Specifies the provisioning model for the node pool.
Behavior for
PROVISIONING_MODEL_UNSPECIFIED:- For CPU node pools: Defaults to
ON_DEMAND. - For GPU node pools:
- If
machine_spec.reservation_affinityis set, it defaults toRESERVATION. - If
machine_spec.reservation_affinityis NOT set, this is an INVALID state and the request will be rejected. An explicitprovisioningModelmust be provided.
- For CPU node pools: Defaults to
- scaling
Spec Property Map - The scaling spec of the node pool.
- startup
Script String - Startup script for the node pool instances.
- zone String
- The zone of the node pool.
AiModelDevelopmentClusterNodePoolBootDisk, AiModelDevelopmentClusterNodePoolBootDiskArgs
- Boot
Disk intSize Gb - Size in GB of the boot disk (default is 100GB).
- Boot
Disk stringType - Type of the boot disk. For non-A3U machines, the default value is "pd-ssd", for A3U machines, the default value is "hyperdisk-balanced". Valid values: "pd-ssd" (Persistent Disk Solid State Drive), "pd-standard" (Persistent Disk Hard Disk Drive) or "hyperdisk-balanced".
- Boot
Disk intSize Gb - Size in GB of the boot disk (default is 100GB).
- Boot
Disk stringType - Type of the boot disk. For non-A3U machines, the default value is "pd-ssd", for A3U machines, the default value is "hyperdisk-balanced". Valid values: "pd-ssd" (Persistent Disk Solid State Drive), "pd-standard" (Persistent Disk Hard Disk Drive) or "hyperdisk-balanced".
- boot_
disk_ numbersize_ gb - Size in GB of the boot disk (default is 100GB).
- boot_
disk_ stringtype - Type of the boot disk. For non-A3U machines, the default value is "pd-ssd", for A3U machines, the default value is "hyperdisk-balanced". Valid values: "pd-ssd" (Persistent Disk Solid State Drive), "pd-standard" (Persistent Disk Hard Disk Drive) or "hyperdisk-balanced".
- boot
Disk IntegerSize Gb - Size in GB of the boot disk (default is 100GB).
- boot
Disk StringType - Type of the boot disk. For non-A3U machines, the default value is "pd-ssd", for A3U machines, the default value is "hyperdisk-balanced". Valid values: "pd-ssd" (Persistent Disk Solid State Drive), "pd-standard" (Persistent Disk Hard Disk Drive) or "hyperdisk-balanced".
- boot
Disk numberSize Gb - Size in GB of the boot disk (default is 100GB).
- boot
Disk stringType - Type of the boot disk. For non-A3U machines, the default value is "pd-ssd", for A3U machines, the default value is "hyperdisk-balanced". Valid values: "pd-ssd" (Persistent Disk Solid State Drive), "pd-standard" (Persistent Disk Hard Disk Drive) or "hyperdisk-balanced".
- boot_
disk_ intsize_ gb - Size in GB of the boot disk (default is 100GB).
- boot_
disk_ strtype - Type of the boot disk. For non-A3U machines, the default value is "pd-ssd", for A3U machines, the default value is "hyperdisk-balanced". Valid values: "pd-ssd" (Persistent Disk Solid State Drive), "pd-standard" (Persistent Disk Hard Disk Drive) or "hyperdisk-balanced".
- boot
Disk NumberSize Gb - Size in GB of the boot disk (default is 100GB).
- boot
Disk StringType - Type of the boot disk. For non-A3U machines, the default value is "pd-ssd", for A3U machines, the default value is "hyperdisk-balanced". Valid values: "pd-ssd" (Persistent Disk Solid State Drive), "pd-standard" (Persistent Disk Hard Disk Drive) or "hyperdisk-balanced".
AiModelDevelopmentClusterNodePoolMachineSpec, AiModelDevelopmentClusterNodePoolMachineSpecArgs
- Machine
Type string The type of the machine.
See the list of machine types supported for prediction
See the list of machine types supported for custom training.
For DeployedModel this field is optional, and the default value is 'n1-standard-2'. For BatchPredictionJob or as part of WorkerPoolSpec this field is required.
- Reservation
Affinity AiModel Development Cluster Node Pool Machine Spec Reservation Affinity - A ReservationAffinity can be used to configure a Vertex AI resource (e.g., a DeployedModel) to draw its Compute Engine resources from a Shared Reservation, or exclusively from on-demand capacity.
- Machine
Type string The type of the machine.
See the list of machine types supported for prediction
See the list of machine types supported for custom training.
For DeployedModel this field is optional, and the default value is 'n1-standard-2'. For BatchPredictionJob or as part of WorkerPoolSpec this field is required.
- Reservation
Affinity AiModel Development Cluster Node Pool Machine Spec Reservation Affinity - A ReservationAffinity can be used to configure a Vertex AI resource (e.g., a DeployedModel) to draw its Compute Engine resources from a Shared Reservation, or exclusively from on-demand capacity.
- machine_
type string The type of the machine.
See the list of machine types supported for prediction
See the list of machine types supported for custom training.
For DeployedModel this field is optional, and the default value is 'n1-standard-2'. For BatchPredictionJob or as part of WorkerPoolSpec this field is required.
- reservation_
affinity object - A ReservationAffinity can be used to configure a Vertex AI resource (e.g., a DeployedModel) to draw its Compute Engine resources from a Shared Reservation, or exclusively from on-demand capacity.
- machine
Type String The type of the machine.
See the list of machine types supported for prediction
See the list of machine types supported for custom training.
For DeployedModel this field is optional, and the default value is 'n1-standard-2'. For BatchPredictionJob or as part of WorkerPoolSpec this field is required.
- reservation
Affinity AiModel Development Cluster Node Pool Machine Spec Reservation Affinity - A ReservationAffinity can be used to configure a Vertex AI resource (e.g., a DeployedModel) to draw its Compute Engine resources from a Shared Reservation, or exclusively from on-demand capacity.
- machine
Type string The type of the machine.
See the list of machine types supported for prediction
See the list of machine types supported for custom training.
For DeployedModel this field is optional, and the default value is 'n1-standard-2'. For BatchPredictionJob or as part of WorkerPoolSpec this field is required.
- reservation
Affinity AiModel Development Cluster Node Pool Machine Spec Reservation Affinity - A ReservationAffinity can be used to configure a Vertex AI resource (e.g., a DeployedModel) to draw its Compute Engine resources from a Shared Reservation, or exclusively from on-demand capacity.
- machine_
type str The type of the machine.
See the list of machine types supported for prediction
See the list of machine types supported for custom training.
For DeployedModel this field is optional, and the default value is 'n1-standard-2'. For BatchPredictionJob or as part of WorkerPoolSpec this field is required.
- reservation_
affinity AiModel Development Cluster Node Pool Machine Spec Reservation Affinity - A ReservationAffinity can be used to configure a Vertex AI resource (e.g., a DeployedModel) to draw its Compute Engine resources from a Shared Reservation, or exclusively from on-demand capacity.
- machine
Type String The type of the machine.
See the list of machine types supported for prediction
See the list of machine types supported for custom training.
For DeployedModel this field is optional, and the default value is 'n1-standard-2'. For BatchPredictionJob or as part of WorkerPoolSpec this field is required.
- reservation
Affinity Property Map - A ReservationAffinity can be used to configure a Vertex AI resource (e.g., a DeployedModel) to draw its Compute Engine resources from a Shared Reservation, or exclusively from on-demand capacity.
AiModelDevelopmentClusterNodePoolMachineSpecReservationAffinity, AiModelDevelopmentClusterNodePoolMachineSpecReservationAffinityArgs
- Reservation
Affinity stringType - Specifies the reservation affinity type. Possible values: NO_RESERVATION ANY_RESERVATION SPECIFIC_RESERVATION
- Values List<string>
- Corresponds to the label values of a reservation resource. This must be the resource name of the reservation, reservation block, or reservation sub- block.
- Reservation
Affinity stringType - Specifies the reservation affinity type. Possible values: NO_RESERVATION ANY_RESERVATION SPECIFIC_RESERVATION
- Values []string
- Corresponds to the label values of a reservation resource. This must be the resource name of the reservation, reservation block, or reservation sub- block.
- reservation_
affinity_ stringtype - Specifies the reservation affinity type. Possible values: NO_RESERVATION ANY_RESERVATION SPECIFIC_RESERVATION
- values list(string)
- Corresponds to the label values of a reservation resource. This must be the resource name of the reservation, reservation block, or reservation sub- block.
- reservation
Affinity StringType - Specifies the reservation affinity type. Possible values: NO_RESERVATION ANY_RESERVATION SPECIFIC_RESERVATION
- values List<String>
- Corresponds to the label values of a reservation resource. This must be the resource name of the reservation, reservation block, or reservation sub- block.
- reservation
Affinity stringType - Specifies the reservation affinity type. Possible values: NO_RESERVATION ANY_RESERVATION SPECIFIC_RESERVATION
- values string[]
- Corresponds to the label values of a reservation resource. This must be the resource name of the reservation, reservation block, or reservation sub- block.
- reservation_
affinity_ strtype - Specifies the reservation affinity type. Possible values: NO_RESERVATION ANY_RESERVATION SPECIFIC_RESERVATION
- values Sequence[str]
- Corresponds to the label values of a reservation resource. This must be the resource name of the reservation, reservation block, or reservation sub- block.
- reservation
Affinity StringType - Specifies the reservation affinity type. Possible values: NO_RESERVATION ANY_RESERVATION SPECIFIC_RESERVATION
- values List<String>
- Corresponds to the label values of a reservation resource. This must be the resource name of the reservation, reservation block, or reservation sub- block.
AiModelDevelopmentClusterNodePoolScalingSpec, AiModelDevelopmentClusterNodePoolScalingSpecArgs
- Max
Node intCount - maximum size of the node pool, must be ≥ min_node_count
- Min
Node intCount - minimum size of the node pool, must be >= 0 and <= max_node_count
- Max
Node intCount - maximum size of the node pool, must be ≥ min_node_count
- Min
Node intCount - minimum size of the node pool, must be >= 0 and <= max_node_count
- max_
node_ numbercount - maximum size of the node pool, must be ≥ min_node_count
- min_
node_ numbercount - minimum size of the node pool, must be >= 0 and <= max_node_count
- max
Node IntegerCount - maximum size of the node pool, must be ≥ min_node_count
- min
Node IntegerCount - minimum size of the node pool, must be >= 0 and <= max_node_count
- max
Node numberCount - maximum size of the node pool, must be ≥ min_node_count
- min
Node numberCount - minimum size of the node pool, must be >= 0 and <= max_node_count
- max_
node_ intcount - maximum size of the node pool, must be ≥ min_node_count
- min_
node_ intcount - minimum size of the node pool, must be >= 0 and <= max_node_count
- max
Node NumberCount - maximum size of the node pool, must be ≥ min_node_count
- min
Node NumberCount - minimum size of the node pool, must be >= 0 and <= max_node_count
AiModelDevelopmentClusterOrchestratorSpec, AiModelDevelopmentClusterOrchestratorSpecArgs
- Kubernetes
Spec AiModel Development Cluster Orchestrator Spec Kubernetes Spec - The GKE orchestrator related spec of the cluster.
- Slurm
Spec AiModel Development Cluster Orchestrator Spec Slurm Spec - The Slurm orchestrator related spec of the cluster.
- Kubernetes
Spec AiModel Development Cluster Orchestrator Spec Kubernetes Spec - The GKE orchestrator related spec of the cluster.
- Slurm
Spec AiModel Development Cluster Orchestrator Spec Slurm Spec - The Slurm orchestrator related spec of the cluster.
- kubernetes_
spec object - The GKE orchestrator related spec of the cluster.
- slurm_
spec object - The Slurm orchestrator related spec of the cluster.
- kubernetes
Spec AiModel Development Cluster Orchestrator Spec Kubernetes Spec - The GKE orchestrator related spec of the cluster.
- slurm
Spec AiModel Development Cluster Orchestrator Spec Slurm Spec - The Slurm orchestrator related spec of the cluster.
- kubernetes
Spec AiModel Development Cluster Orchestrator Spec Kubernetes Spec - The GKE orchestrator related spec of the cluster.
- slurm
Spec AiModel Development Cluster Orchestrator Spec Slurm Spec - The Slurm orchestrator related spec of the cluster.
- kubernetes_
spec AiModel Development Cluster Orchestrator Spec Kubernetes Spec - The GKE orchestrator related spec of the cluster.
- slurm_
spec AiModel Development Cluster Orchestrator Spec Slurm Spec - The Slurm orchestrator related spec of the cluster.
- kubernetes
Spec Property Map - The GKE orchestrator related spec of the cluster.
- slurm
Spec Property Map - The Slurm orchestrator related spec of the cluster.
AiModelDevelopmentClusterOrchestratorSpecKubernetesSpec, AiModelDevelopmentClusterOrchestratorSpecKubernetesSpecArgs
- Control
Plane stringVersion - The Kubernetes version of the GKE control plane. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to use the service default version. Upgrades can advance at most one minor version at a time; downgrades are rejected.
- Enable
Ip boolAlias - Whether to enable IP alias.
- Node
Pool List<string>Ids - IDs of the NodePools to be included in this GKE cluster. These IDs must correspond to NodePools defined in the 'node_pools' field.
- Node
Pool Dictionary<string, string>Versions - The Kubernetes versions for specific node pools, keyed by node pool ID. When specified, all node pool IDs in the cluster must be included. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to align with the control plane version. Node pools cannot be newer than the control plane or more than 2 minor versions behind it; upgrades can advance at most one minor version at a time and downgrades are rejected.
- Public
Access boolEndpoint Enabled - Whether the GKE cluster's control plane endpoint is publicly accessible. If true, the cluster is created with a public access endpoint (i.e. the private cluster configuration is disabled). If false or unset, the cluster is created with only private control plane access by default.
- Control
Plane stringVersion - The Kubernetes version of the GKE control plane. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to use the service default version. Upgrades can advance at most one minor version at a time; downgrades are rejected.
- Enable
Ip boolAlias - Whether to enable IP alias.
- Node
Pool []stringIds - IDs of the NodePools to be included in this GKE cluster. These IDs must correspond to NodePools defined in the 'node_pools' field.
- Node
Pool map[string]stringVersions - The Kubernetes versions for specific node pools, keyed by node pool ID. When specified, all node pool IDs in the cluster must be included. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to align with the control plane version. Node pools cannot be newer than the control plane or more than 2 minor versions behind it; upgrades can advance at most one minor version at a time and downgrades are rejected.
- Public
Access boolEndpoint Enabled - Whether the GKE cluster's control plane endpoint is publicly accessible. If true, the cluster is created with a public access endpoint (i.e. the private cluster configuration is disabled). If false or unset, the cluster is created with only private control plane access by default.
- control_
plane_ stringversion - The Kubernetes version of the GKE control plane. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to use the service default version. Upgrades can advance at most one minor version at a time; downgrades are rejected.
- enable_
ip_ boolalias - Whether to enable IP alias.
- node_
pool_ list(string)ids - IDs of the NodePools to be included in this GKE cluster. These IDs must correspond to NodePools defined in the 'node_pools' field.
- node_
pool_ map(string)versions - The Kubernetes versions for specific node pools, keyed by node pool ID. When specified, all node pool IDs in the cluster must be included. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to align with the control plane version. Node pools cannot be newer than the control plane or more than 2 minor versions behind it; upgrades can advance at most one minor version at a time and downgrades are rejected.
- public_
access_ boolendpoint_ enabled - Whether the GKE cluster's control plane endpoint is publicly accessible. If true, the cluster is created with a public access endpoint (i.e. the private cluster configuration is disabled). If false or unset, the cluster is created with only private control plane access by default.
- control
Plane StringVersion - The Kubernetes version of the GKE control plane. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to use the service default version. Upgrades can advance at most one minor version at a time; downgrades are rejected.
- enable
Ip BooleanAlias - Whether to enable IP alias.
- node
Pool List<String>Ids - IDs of the NodePools to be included in this GKE cluster. These IDs must correspond to NodePools defined in the 'node_pools' field.
- node
Pool Map<String,String>Versions - The Kubernetes versions for specific node pools, keyed by node pool ID. When specified, all node pool IDs in the cluster must be included. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to align with the control plane version. Node pools cannot be newer than the control plane or more than 2 minor versions behind it; upgrades can advance at most one minor version at a time and downgrades are rejected.
- public
Access BooleanEndpoint Enabled - Whether the GKE cluster's control plane endpoint is publicly accessible. If true, the cluster is created with a public access endpoint (i.e. the private cluster configuration is disabled). If false or unset, the cluster is created with only private control plane access by default.
- control
Plane stringVersion - The Kubernetes version of the GKE control plane. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to use the service default version. Upgrades can advance at most one minor version at a time; downgrades are rejected.
- enable
Ip booleanAlias - Whether to enable IP alias.
- node
Pool string[]Ids - IDs of the NodePools to be included in this GKE cluster. These IDs must correspond to NodePools defined in the 'node_pools' field.
- node
Pool {[key: string]: string}Versions - The Kubernetes versions for specific node pools, keyed by node pool ID. When specified, all node pool IDs in the cluster must be included. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to align with the control plane version. Node pools cannot be newer than the control plane or more than 2 minor versions behind it; upgrades can advance at most one minor version at a time and downgrades are rejected.
- public
Access booleanEndpoint Enabled - Whether the GKE cluster's control plane endpoint is publicly accessible. If true, the cluster is created with a public access endpoint (i.e. the private cluster configuration is disabled). If false or unset, the cluster is created with only private control plane access by default.
- control_
plane_ strversion - The Kubernetes version of the GKE control plane. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to use the service default version. Upgrades can advance at most one minor version at a time; downgrades are rejected.
- enable_
ip_ boolalias - Whether to enable IP alias.
- node_
pool_ Sequence[str]ids - IDs of the NodePools to be included in this GKE cluster. These IDs must correspond to NodePools defined in the 'node_pools' field.
- node_
pool_ Mapping[str, str]versions - The Kubernetes versions for specific node pools, keyed by node pool ID. When specified, all node pool IDs in the cluster must be included. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to align with the control plane version. Node pools cannot be newer than the control plane or more than 2 minor versions behind it; upgrades can advance at most one minor version at a time and downgrades are rejected.
- public_
access_ boolendpoint_ enabled - Whether the GKE cluster's control plane endpoint is publicly accessible. If true, the cluster is created with a public access endpoint (i.e. the private cluster configuration is disabled). If false or unset, the cluster is created with only private control plane access by default.
- control
Plane StringVersion - The Kubernetes version of the GKE control plane. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to use the service default version. Upgrades can advance at most one minor version at a time; downgrades are rejected.
- enable
Ip BooleanAlias - Whether to enable IP alias.
- node
Pool List<String>Ids - IDs of the NodePools to be included in this GKE cluster. These IDs must correspond to NodePools defined in the 'node_pools' field.
- node
Pool Map<String>Versions - The Kubernetes versions for specific node pools, keyed by node pool ID. When specified, all node pool IDs in the cluster must be included. Accepted formats are minor version (for example, '"1.30"'), patch version (for example, '"1.30.5"'), full GKE release version (for example, '"1.30.5-gke.1014001"'), or '"-"' to align with the control plane version. Node pools cannot be newer than the control plane or more than 2 minor versions behind it; upgrades can advance at most one minor version at a time and downgrades are rejected.
- public
Access BooleanEndpoint Enabled - Whether the GKE cluster's control plane endpoint is publicly accessible. If true, the cluster is created with a public access endpoint (i.e. the private cluster configuration is disabled). If false or unset, the cluster is created with only private control plane access by default.
AiModelDevelopmentClusterOrchestratorSpecSlurmSpec, AiModelDevelopmentClusterOrchestratorSpecSlurmSpecArgs
- Home
Directory stringStorage - Full name of the Filestore to be mounted as /home directory on every node.
- Login
Node stringPool Id - The node pool id to be used as slurm login node.
- Config Dictionary<string, string>
Cluster-scoped Slurm settings, emitted as global 'Key=Value' lines in slurm.conf.
Keys are Slurm parameter names exactly as documented by SchedMD (https://slurm.schedmd.com/slurm.conf.html), matched case-insensitively. Values are written the way slurm.conf writes them, including composite forms such as 'bf_window=1440,bf_resolution=600'.
Example: { "PriorityType": "priority/multifactor", "PriorityWeightAge": "1000", "PreemptMode": "SUSPEND,GANG", "SchedulerParameters": "bf_window=1440,bf_resolution=600" }
Unsupported, reserved, or misspelled parameters, and values that are not legal for the parameter, are rejected at request time with the offending parameter named; they are never silently dropped. Settings that apply to a single partition or node pool belong in 'SlurmPartition.config' or 'SlurmNodeSet.config' instead - several parameters (e.g. 'DefMemPerCPU', 'OverTimeLimit', 'PreemptMode') are valid at more than one scope with different meanings and different legal values.
On update this map is replaced wholesale; supply the complete desired set.
- Default
Partition stringId Id of the partition that Slurm jobs submitted without an explicit partition run on, for example 'sbatch' without '--partition'. Must match the 'id' of one of 'partitions'.
If unset, the service uses the first entry of 'partitions' and this field stays empty; the cluster always has a default partition whether or not this field is set. Setting it makes the choice independent of list order, which matters because 'partitions' is replaced wholesale on update: reordering the list, or prepending a partition to it, otherwise moves the default and silently redirects every job submitted without '--partition'.
- Epilog
Bash List<string>Scripts - Custom bash scripts executed after each slurm job on every allocated node.
- Node
Sets List<AiModel Development Cluster Orchestrator Spec Slurm Spec Node Set> - Node-scoped Slurm settings, keyed by node pool. At most one entry per node pool. Node pools without an entry use the settings derived by the cluster's Slurm controller.
- Partitions
List<Ai
Model Development Cluster Orchestrator Spec Slurm Spec Partition> - The Slurm partitions spec.
- Prolog
Bash List<string>Scripts - Custom bash scripts executed before each slurm job on every allocated node.
- Home
Directory stringStorage - Full name of the Filestore to be mounted as /home directory on every node.
- Login
Node stringPool Id - The node pool id to be used as slurm login node.
- Config map[string]string
Cluster-scoped Slurm settings, emitted as global 'Key=Value' lines in slurm.conf.
Keys are Slurm parameter names exactly as documented by SchedMD (https://slurm.schedmd.com/slurm.conf.html), matched case-insensitively. Values are written the way slurm.conf writes them, including composite forms such as 'bf_window=1440,bf_resolution=600'.
Example: { "PriorityType": "priority/multifactor", "PriorityWeightAge": "1000", "PreemptMode": "SUSPEND,GANG", "SchedulerParameters": "bf_window=1440,bf_resolution=600" }
Unsupported, reserved, or misspelled parameters, and values that are not legal for the parameter, are rejected at request time with the offending parameter named; they are never silently dropped. Settings that apply to a single partition or node pool belong in 'SlurmPartition.config' or 'SlurmNodeSet.config' instead - several parameters (e.g. 'DefMemPerCPU', 'OverTimeLimit', 'PreemptMode') are valid at more than one scope with different meanings and different legal values.
On update this map is replaced wholesale; supply the complete desired set.
- Default
Partition stringId Id of the partition that Slurm jobs submitted without an explicit partition run on, for example 'sbatch' without '--partition'. Must match the 'id' of one of 'partitions'.
If unset, the service uses the first entry of 'partitions' and this field stays empty; the cluster always has a default partition whether or not this field is set. Setting it makes the choice independent of list order, which matters because 'partitions' is replaced wholesale on update: reordering the list, or prepending a partition to it, otherwise moves the default and silently redirects every job submitted without '--partition'.
- Epilog
Bash []stringScripts - Custom bash scripts executed after each slurm job on every allocated node.
- Node
Sets []AiModel Development Cluster Orchestrator Spec Slurm Spec Node Set - Node-scoped Slurm settings, keyed by node pool. At most one entry per node pool. Node pools without an entry use the settings derived by the cluster's Slurm controller.
- Partitions
[]Ai
Model Development Cluster Orchestrator Spec Slurm Spec Partition - The Slurm partitions spec.
- Prolog
Bash []stringScripts - Custom bash scripts executed before each slurm job on every allocated node.
- home_
directory_ stringstorage - Full name of the Filestore to be mounted as /home directory on every node.
- login_
node_ stringpool_ id - The node pool id to be used as slurm login node.
- config map(string)
Cluster-scoped Slurm settings, emitted as global 'Key=Value' lines in slurm.conf.
Keys are Slurm parameter names exactly as documented by SchedMD (https://slurm.schedmd.com/slurm.conf.html), matched case-insensitively. Values are written the way slurm.conf writes them, including composite forms such as 'bf_window=1440,bf_resolution=600'.
Example: { "PriorityType": "priority/multifactor", "PriorityWeightAge": "1000", "PreemptMode": "SUSPEND,GANG", "SchedulerParameters": "bf_window=1440,bf_resolution=600" }
Unsupported, reserved, or misspelled parameters, and values that are not legal for the parameter, are rejected at request time with the offending parameter named; they are never silently dropped. Settings that apply to a single partition or node pool belong in 'SlurmPartition.config' or 'SlurmNodeSet.config' instead - several parameters (e.g. 'DefMemPerCPU', 'OverTimeLimit', 'PreemptMode') are valid at more than one scope with different meanings and different legal values.
On update this map is replaced wholesale; supply the complete desired set.
- default_
partition_ stringid Id of the partition that Slurm jobs submitted without an explicit partition run on, for example 'sbatch' without '--partition'. Must match the 'id' of one of 'partitions'.
If unset, the service uses the first entry of 'partitions' and this field stays empty; the cluster always has a default partition whether or not this field is set. Setting it makes the choice independent of list order, which matters because 'partitions' is replaced wholesale on update: reordering the list, or prepending a partition to it, otherwise moves the default and silently redirects every job submitted without '--partition'.
- epilog_
bash_ list(string)scripts - Custom bash scripts executed after each slurm job on every allocated node.
- node_
sets list(object) - Node-scoped Slurm settings, keyed by node pool. At most one entry per node pool. Node pools without an entry use the settings derived by the cluster's Slurm controller.
- partitions list(object)
- The Slurm partitions spec.
- prolog_
bash_ list(string)scripts - Custom bash scripts executed before each slurm job on every allocated node.
- home
Directory StringStorage - Full name of the Filestore to be mounted as /home directory on every node.
- login
Node StringPool Id - The node pool id to be used as slurm login node.
- config Map<String,String>
Cluster-scoped Slurm settings, emitted as global 'Key=Value' lines in slurm.conf.
Keys are Slurm parameter names exactly as documented by SchedMD (https://slurm.schedmd.com/slurm.conf.html), matched case-insensitively. Values are written the way slurm.conf writes them, including composite forms such as 'bf_window=1440,bf_resolution=600'.
Example: { "PriorityType": "priority/multifactor", "PriorityWeightAge": "1000", "PreemptMode": "SUSPEND,GANG", "SchedulerParameters": "bf_window=1440,bf_resolution=600" }
Unsupported, reserved, or misspelled parameters, and values that are not legal for the parameter, are rejected at request time with the offending parameter named; they are never silently dropped. Settings that apply to a single partition or node pool belong in 'SlurmPartition.config' or 'SlurmNodeSet.config' instead - several parameters (e.g. 'DefMemPerCPU', 'OverTimeLimit', 'PreemptMode') are valid at more than one scope with different meanings and different legal values.
On update this map is replaced wholesale; supply the complete desired set.
- default
Partition StringId Id of the partition that Slurm jobs submitted without an explicit partition run on, for example 'sbatch' without '--partition'. Must match the 'id' of one of 'partitions'.
If unset, the service uses the first entry of 'partitions' and this field stays empty; the cluster always has a default partition whether or not this field is set. Setting it makes the choice independent of list order, which matters because 'partitions' is replaced wholesale on update: reordering the list, or prepending a partition to it, otherwise moves the default and silently redirects every job submitted without '--partition'.
- epilog
Bash List<String>Scripts - Custom bash scripts executed after each slurm job on every allocated node.
- node
Sets List<AiModel Development Cluster Orchestrator Spec Slurm Spec Node Set> - Node-scoped Slurm settings, keyed by node pool. At most one entry per node pool. Node pools without an entry use the settings derived by the cluster's Slurm controller.
- partitions
List<Ai
Model Development Cluster Orchestrator Spec Slurm Spec Partition> - The Slurm partitions spec.
- prolog
Bash List<String>Scripts - Custom bash scripts executed before each slurm job on every allocated node.
- home
Directory stringStorage - Full name of the Filestore to be mounted as /home directory on every node.
- login
Node stringPool Id - The node pool id to be used as slurm login node.
- config {[key: string]: string}
Cluster-scoped Slurm settings, emitted as global 'Key=Value' lines in slurm.conf.
Keys are Slurm parameter names exactly as documented by SchedMD (https://slurm.schedmd.com/slurm.conf.html), matched case-insensitively. Values are written the way slurm.conf writes them, including composite forms such as 'bf_window=1440,bf_resolution=600'.
Example: { "PriorityType": "priority/multifactor", "PriorityWeightAge": "1000", "PreemptMode": "SUSPEND,GANG", "SchedulerParameters": "bf_window=1440,bf_resolution=600" }
Unsupported, reserved, or misspelled parameters, and values that are not legal for the parameter, are rejected at request time with the offending parameter named; they are never silently dropped. Settings that apply to a single partition or node pool belong in 'SlurmPartition.config' or 'SlurmNodeSet.config' instead - several parameters (e.g. 'DefMemPerCPU', 'OverTimeLimit', 'PreemptMode') are valid at more than one scope with different meanings and different legal values.
On update this map is replaced wholesale; supply the complete desired set.
- default
Partition stringId Id of the partition that Slurm jobs submitted without an explicit partition run on, for example 'sbatch' without '--partition'. Must match the 'id' of one of 'partitions'.
If unset, the service uses the first entry of 'partitions' and this field stays empty; the cluster always has a default partition whether or not this field is set. Setting it makes the choice independent of list order, which matters because 'partitions' is replaced wholesale on update: reordering the list, or prepending a partition to it, otherwise moves the default and silently redirects every job submitted without '--partition'.
- epilog
Bash string[]Scripts - Custom bash scripts executed after each slurm job on every allocated node.
- node
Sets AiModel Development Cluster Orchestrator Spec Slurm Spec Node Set[] - Node-scoped Slurm settings, keyed by node pool. At most one entry per node pool. Node pools without an entry use the settings derived by the cluster's Slurm controller.
- partitions
Ai
Model Development Cluster Orchestrator Spec Slurm Spec Partition[] - The Slurm partitions spec.
- prolog
Bash string[]Scripts - Custom bash scripts executed before each slurm job on every allocated node.
- home_
directory_ strstorage - Full name of the Filestore to be mounted as /home directory on every node.
- login_
node_ strpool_ id - The node pool id to be used as slurm login node.
- config Mapping[str, str]
Cluster-scoped Slurm settings, emitted as global 'Key=Value' lines in slurm.conf.
Keys are Slurm parameter names exactly as documented by SchedMD (https://slurm.schedmd.com/slurm.conf.html), matched case-insensitively. Values are written the way slurm.conf writes them, including composite forms such as 'bf_window=1440,bf_resolution=600'.
Example: { "PriorityType": "priority/multifactor", "PriorityWeightAge": "1000", "PreemptMode": "SUSPEND,GANG", "SchedulerParameters": "bf_window=1440,bf_resolution=600" }
Unsupported, reserved, or misspelled parameters, and values that are not legal for the parameter, are rejected at request time with the offending parameter named; they are never silently dropped. Settings that apply to a single partition or node pool belong in 'SlurmPartition.config' or 'SlurmNodeSet.config' instead - several parameters (e.g. 'DefMemPerCPU', 'OverTimeLimit', 'PreemptMode') are valid at more than one scope with different meanings and different legal values.
On update this map is replaced wholesale; supply the complete desired set.
- default_
partition_ strid Id of the partition that Slurm jobs submitted without an explicit partition run on, for example 'sbatch' without '--partition'. Must match the 'id' of one of 'partitions'.
If unset, the service uses the first entry of 'partitions' and this field stays empty; the cluster always has a default partition whether or not this field is set. Setting it makes the choice independent of list order, which matters because 'partitions' is replaced wholesale on update: reordering the list, or prepending a partition to it, otherwise moves the default and silently redirects every job submitted without '--partition'.
- epilog_
bash_ Sequence[str]scripts - Custom bash scripts executed after each slurm job on every allocated node.
- node_
sets Sequence[AiModel Development Cluster Orchestrator Spec Slurm Spec Node Set] - Node-scoped Slurm settings, keyed by node pool. At most one entry per node pool. Node pools without an entry use the settings derived by the cluster's Slurm controller.
- partitions
Sequence[Ai
Model Development Cluster Orchestrator Spec Slurm Spec Partition] - The Slurm partitions spec.
- prolog_
bash_ Sequence[str]scripts - Custom bash scripts executed before each slurm job on every allocated node.
- home
Directory StringStorage - Full name of the Filestore to be mounted as /home directory on every node.
- login
Node StringPool Id - The node pool id to be used as slurm login node.
- config Map<String>
Cluster-scoped Slurm settings, emitted as global 'Key=Value' lines in slurm.conf.
Keys are Slurm parameter names exactly as documented by SchedMD (https://slurm.schedmd.com/slurm.conf.html), matched case-insensitively. Values are written the way slurm.conf writes them, including composite forms such as 'bf_window=1440,bf_resolution=600'.
Example: { "PriorityType": "priority/multifactor", "PriorityWeightAge": "1000", "PreemptMode": "SUSPEND,GANG", "SchedulerParameters": "bf_window=1440,bf_resolution=600" }
Unsupported, reserved, or misspelled parameters, and values that are not legal for the parameter, are rejected at request time with the offending parameter named; they are never silently dropped. Settings that apply to a single partition or node pool belong in 'SlurmPartition.config' or 'SlurmNodeSet.config' instead - several parameters (e.g. 'DefMemPerCPU', 'OverTimeLimit', 'PreemptMode') are valid at more than one scope with different meanings and different legal values.
On update this map is replaced wholesale; supply the complete desired set.
- default
Partition StringId Id of the partition that Slurm jobs submitted without an explicit partition run on, for example 'sbatch' without '--partition'. Must match the 'id' of one of 'partitions'.
If unset, the service uses the first entry of 'partitions' and this field stays empty; the cluster always has a default partition whether or not this field is set. Setting it makes the choice independent of list order, which matters because 'partitions' is replaced wholesale on update: reordering the list, or prepending a partition to it, otherwise moves the default and silently redirects every job submitted without '--partition'.
- epilog
Bash List<String>Scripts - Custom bash scripts executed after each slurm job on every allocated node.
- node
Sets List<Property Map> - Node-scoped Slurm settings, keyed by node pool. At most one entry per node pool. Node pools without an entry use the settings derived by the cluster's Slurm controller.
- partitions List<Property Map>
- The Slurm partitions spec.
- prolog
Bash List<String>Scripts - Custom bash scripts executed before each slurm job on every allocated node.
AiModelDevelopmentClusterOrchestratorSpecSlurmSpecNodeSet, AiModelDevelopmentClusterOrchestratorSpecSlurmSpecNodeSetArgs
- Node
Pool stringId - (Output) The ID of the node pool.
- Config Dictionary<string, string>
Node-scoped Slurm settings, emitted as additional 'Key=Value' pairs on the 'NodeName=' line for this node pool's nodes in slurm.conf.
Example: { "Features": "a3mega,nvlink", "Weight": "10", "MemSpecLimit": "4096" }
On update this map is replaced wholesale; supply the complete desired set.
- Node
Pool stringId - (Output) The ID of the node pool.
- Config map[string]string
Node-scoped Slurm settings, emitted as additional 'Key=Value' pairs on the 'NodeName=' line for this node pool's nodes in slurm.conf.
Example: { "Features": "a3mega,nvlink", "Weight": "10", "MemSpecLimit": "4096" }
On update this map is replaced wholesale; supply the complete desired set.
- node_
pool_ stringid - (Output) The ID of the node pool.
- config map(string)
Node-scoped Slurm settings, emitted as additional 'Key=Value' pairs on the 'NodeName=' line for this node pool's nodes in slurm.conf.
Example: { "Features": "a3mega,nvlink", "Weight": "10", "MemSpecLimit": "4096" }
On update this map is replaced wholesale; supply the complete desired set.
- node
Pool StringId - (Output) The ID of the node pool.
- config Map<String,String>
Node-scoped Slurm settings, emitted as additional 'Key=Value' pairs on the 'NodeName=' line for this node pool's nodes in slurm.conf.
Example: { "Features": "a3mega,nvlink", "Weight": "10", "MemSpecLimit": "4096" }
On update this map is replaced wholesale; supply the complete desired set.
- node
Pool stringId - (Output) The ID of the node pool.
- config {[key: string]: string}
Node-scoped Slurm settings, emitted as additional 'Key=Value' pairs on the 'NodeName=' line for this node pool's nodes in slurm.conf.
Example: { "Features": "a3mega,nvlink", "Weight": "10", "MemSpecLimit": "4096" }
On update this map is replaced wholesale; supply the complete desired set.
- node_
pool_ strid - (Output) The ID of the node pool.
- config Mapping[str, str]
Node-scoped Slurm settings, emitted as additional 'Key=Value' pairs on the 'NodeName=' line for this node pool's nodes in slurm.conf.
Example: { "Features": "a3mega,nvlink", "Weight": "10", "MemSpecLimit": "4096" }
On update this map is replaced wholesale; supply the complete desired set.
- node
Pool StringId - (Output) The ID of the node pool.
- config Map<String>
Node-scoped Slurm settings, emitted as additional 'Key=Value' pairs on the 'NodeName=' line for this node pool's nodes in slurm.conf.
Example: { "Features": "a3mega,nvlink", "Weight": "10", "MemSpecLimit": "4096" }
On update this map is replaced wholesale; supply the complete desired set.
AiModelDevelopmentClusterOrchestratorSpecSlurmSpecPartition, AiModelDevelopmentClusterOrchestratorSpecSlurmSpecPartitionArgs
- Id string
- an identifier for the resource with format
projects/{{project}}/locations/{{region}}/modelDevelopmentClusters/{{model_development_cluster_id}} - Node
Pool List<string>Ids - Node pools for the slurm partition
- Config Dictionary<string, string>
Partition-scoped Slurm settings, emitted as additional 'Key=Value' pairs on this partition's 'PartitionName=' line in slurm.conf.
Example: { "MaxTime": "4-00:00:00", "DenyAccounts": "low-pri", "PriorityTier": "20", "OverSubscribe": "YES:4" }
Note: values set here take precedence over service-computed defaults for the same parameter (for example 'OverSubscribe' and 'DefMemPerCPU', which the service otherwise derives from the partition's exclusivity and from the memory per CPU of its nodes).
'Default' is rejected here: which partition is the default is a cluster-scoped choice, so it belongs in 'SlurmSpec.default_partition_id'.
- Id string
- an identifier for the resource with format
projects/{{project}}/locations/{{region}}/modelDevelopmentClusters/{{model_development_cluster_id}} - Node
Pool []stringIds - Node pools for the slurm partition
- Config map[string]string
Partition-scoped Slurm settings, emitted as additional 'Key=Value' pairs on this partition's 'PartitionName=' line in slurm.conf.
Example: { "MaxTime": "4-00:00:00", "DenyAccounts": "low-pri", "PriorityTier": "20", "OverSubscribe": "YES:4" }
Note: values set here take precedence over service-computed defaults for the same parameter (for example 'OverSubscribe' and 'DefMemPerCPU', which the service otherwise derives from the partition's exclusivity and from the memory per CPU of its nodes).
'Default' is rejected here: which partition is the default is a cluster-scoped choice, so it belongs in 'SlurmSpec.default_partition_id'.
- id string
- an identifier for the resource with format
projects/{{project}}/locations/{{region}}/modelDevelopmentClusters/{{model_development_cluster_id}} - node_
pool_ list(string)ids - Node pools for the slurm partition
- config map(string)
Partition-scoped Slurm settings, emitted as additional 'Key=Value' pairs on this partition's 'PartitionName=' line in slurm.conf.
Example: { "MaxTime": "4-00:00:00", "DenyAccounts": "low-pri", "PriorityTier": "20", "OverSubscribe": "YES:4" }
Note: values set here take precedence over service-computed defaults for the same parameter (for example 'OverSubscribe' and 'DefMemPerCPU', which the service otherwise derives from the partition's exclusivity and from the memory per CPU of its nodes).
'Default' is rejected here: which partition is the default is a cluster-scoped choice, so it belongs in 'SlurmSpec.default_partition_id'.
- id String
- an identifier for the resource with format
projects/{{project}}/locations/{{region}}/modelDevelopmentClusters/{{model_development_cluster_id}} - node
Pool List<String>Ids - Node pools for the slurm partition
- config Map<String,String>
Partition-scoped Slurm settings, emitted as additional 'Key=Value' pairs on this partition's 'PartitionName=' line in slurm.conf.
Example: { "MaxTime": "4-00:00:00", "DenyAccounts": "low-pri", "PriorityTier": "20", "OverSubscribe": "YES:4" }
Note: values set here take precedence over service-computed defaults for the same parameter (for example 'OverSubscribe' and 'DefMemPerCPU', which the service otherwise derives from the partition's exclusivity and from the memory per CPU of its nodes).
'Default' is rejected here: which partition is the default is a cluster-scoped choice, so it belongs in 'SlurmSpec.default_partition_id'.
- id string
- an identifier for the resource with format
projects/{{project}}/locations/{{region}}/modelDevelopmentClusters/{{model_development_cluster_id}} - node
Pool string[]Ids - Node pools for the slurm partition
- config {[key: string]: string}
Partition-scoped Slurm settings, emitted as additional 'Key=Value' pairs on this partition's 'PartitionName=' line in slurm.conf.
Example: { "MaxTime": "4-00:00:00", "DenyAccounts": "low-pri", "PriorityTier": "20", "OverSubscribe": "YES:4" }
Note: values set here take precedence over service-computed defaults for the same parameter (for example 'OverSubscribe' and 'DefMemPerCPU', which the service otherwise derives from the partition's exclusivity and from the memory per CPU of its nodes).
'Default' is rejected here: which partition is the default is a cluster-scoped choice, so it belongs in 'SlurmSpec.default_partition_id'.
- id str
- an identifier for the resource with format
projects/{{project}}/locations/{{region}}/modelDevelopmentClusters/{{model_development_cluster_id}} - node_
pool_ Sequence[str]ids - Node pools for the slurm partition
- config Mapping[str, str]
Partition-scoped Slurm settings, emitted as additional 'Key=Value' pairs on this partition's 'PartitionName=' line in slurm.conf.
Example: { "MaxTime": "4-00:00:00", "DenyAccounts": "low-pri", "PriorityTier": "20", "OverSubscribe": "YES:4" }
Note: values set here take precedence over service-computed defaults for the same parameter (for example 'OverSubscribe' and 'DefMemPerCPU', which the service otherwise derives from the partition's exclusivity and from the memory per CPU of its nodes).
'Default' is rejected here: which partition is the default is a cluster-scoped choice, so it belongs in 'SlurmSpec.default_partition_id'.
- id String
- an identifier for the resource with format
projects/{{project}}/locations/{{region}}/modelDevelopmentClusters/{{model_development_cluster_id}} - node
Pool List<String>Ids - Node pools for the slurm partition
- config Map<String>
Partition-scoped Slurm settings, emitted as additional 'Key=Value' pairs on this partition's 'PartitionName=' line in slurm.conf.
Example: { "MaxTime": "4-00:00:00", "DenyAccounts": "low-pri", "PriorityTier": "20", "OverSubscribe": "YES:4" }
Note: values set here take precedence over service-computed defaults for the same parameter (for example 'OverSubscribe' and 'DefMemPerCPU', which the service otherwise derives from the partition's exclusivity and from the memory per CPU of its nodes).
'Default' is rejected here: which partition is the default is a cluster-scoped choice, so it belongs in 'SlurmSpec.default_partition_id'.
AiModelDevelopmentClusterRuntime, AiModelDevelopmentClusterRuntimeArgs
- Training
Services List<AiModel Development Cluster Runtime Training Service> - (Output) The training services running on this Vertex AI Training Cluster. Structure is documented below.
- Xmanagers
List<Ai
Model Development Cluster Runtime Xmanager> - (Output) A purpose-built, output-only view of the XManager instance attached to a cluster. Structure is documented below.
- Training
Services []AiModel Development Cluster Runtime Training Service - (Output) The training services running on this Vertex AI Training Cluster. Structure is documented below.
- Xmanagers
[]Ai
Model Development Cluster Runtime Xmanager - (Output) A purpose-built, output-only view of the XManager instance attached to a cluster. Structure is documented below.
- training_
services list(object) - (Output) The training services running on this Vertex AI Training Cluster. Structure is documented below.
- xmanagers list(object)
- (Output) A purpose-built, output-only view of the XManager instance attached to a cluster. Structure is documented below.
- training
Services List<AiModel Development Cluster Runtime Training Service> - (Output) The training services running on this Vertex AI Training Cluster. Structure is documented below.
- xmanagers
List<Ai
Model Development Cluster Runtime Xmanager> - (Output) A purpose-built, output-only view of the XManager instance attached to a cluster. Structure is documented below.
- training
Services AiModel Development Cluster Runtime Training Service[] - (Output) The training services running on this Vertex AI Training Cluster. Structure is documented below.
- xmanagers
Ai
Model Development Cluster Runtime Xmanager[] - (Output) A purpose-built, output-only view of the XManager instance attached to a cluster. Structure is documented below.
- training_
services Sequence[AiModel Development Cluster Runtime Training Service] - (Output) The training services running on this Vertex AI Training Cluster. Structure is documented below.
- xmanagers
Sequence[Ai
Model Development Cluster Runtime Xmanager] - (Output) A purpose-built, output-only view of the XManager instance attached to a cluster. Structure is documented below.
- training
Services List<Property Map> - (Output) The training services running on this Vertex AI Training Cluster. Structure is documented below.
- xmanagers List<Property Map>
- (Output) A purpose-built, output-only view of the XManager instance attached to a cluster. Structure is documented below.
AiModelDevelopmentClusterRuntimeSpec, AiModelDevelopmentClusterRuntimeSpecArgs
- Service
Account string - The default service account used by the cluster when running any workloads on it. If not specified, the default compute engine service account will be used.
- Service
Account string - The default service account used by the cluster when running any workloads on it. If not specified, the default compute engine service account will be used.
- service_
account string - The default service account used by the cluster when running any workloads on it. If not specified, the default compute engine service account will be used.
- service
Account String - The default service account used by the cluster when running any workloads on it. If not specified, the default compute engine service account will be used.
- service
Account string - The default service account used by the cluster when running any workloads on it. If not specified, the default compute engine service account will be used.
- service_
account str - The default service account used by the cluster when running any workloads on it. If not specified, the default compute engine service account will be used.
- service
Account String - The default service account used by the cluster when running any workloads on it. If not specified, the default compute engine service account will be used.
AiModelDevelopmentClusterRuntimeTrainingService, AiModelDevelopmentClusterRuntimeTrainingServiceArgs
- Create
Time string - (Output) Time when the TrainingService was created.
- Deployment
Specs List<AiModel Development Cluster Runtime Training Service Deployment Spec> - (Output) Orchestrator-agnostic deployment spec. Structure is documented below.
- Gcs
Bucket string - (Output) Cloud Storage bucket name dynamically GCSFuse-mounted on all nodes, for storing checkpoints and other service outputs.
- Image
Uri string - (Output) Cloud Storage URI of the squashfs (.sqsh) container image.
- Name string
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - Service
Statuses List<AiModel Development Cluster Runtime Training Service Service Status> - (Output) Status of a TrainingService. Structure is documented below.
- Update
Time string - (Output) Time when the TrainingService was most recently updated.
- Create
Time string - (Output) Time when the TrainingService was created.
- Deployment
Specs []AiModel Development Cluster Runtime Training Service Deployment Spec - (Output) Orchestrator-agnostic deployment spec. Structure is documented below.
- Gcs
Bucket string - (Output) Cloud Storage bucket name dynamically GCSFuse-mounted on all nodes, for storing checkpoints and other service outputs.
- Image
Uri string - (Output) Cloud Storage URI of the squashfs (.sqsh) container image.
- Name string
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - Service
Statuses []AiModel Development Cluster Runtime Training Service Service Status - (Output) Status of a TrainingService. Structure is documented below.
- Update
Time string - (Output) Time when the TrainingService was most recently updated.
- create_
time string - (Output) Time when the TrainingService was created.
- deployment_
specs list(object) - (Output) Orchestrator-agnostic deployment spec. Structure is documented below.
- gcs_
bucket string - (Output) Cloud Storage bucket name dynamically GCSFuse-mounted on all nodes, for storing checkpoints and other service outputs.
- image_
uri string - (Output) Cloud Storage URI of the squashfs (.sqsh) container image.
- name string
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - service_
statuses list(object) - (Output) Status of a TrainingService. Structure is documented below.
- update_
time string - (Output) Time when the TrainingService was most recently updated.
- create
Time String - (Output) Time when the TrainingService was created.
- deployment
Specs List<AiModel Development Cluster Runtime Training Service Deployment Spec> - (Output) Orchestrator-agnostic deployment spec. Structure is documented below.
- gcs
Bucket String - (Output) Cloud Storage bucket name dynamically GCSFuse-mounted on all nodes, for storing checkpoints and other service outputs.
- image
Uri String - (Output) Cloud Storage URI of the squashfs (.sqsh) container image.
- name String
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - service
Statuses List<AiModel Development Cluster Runtime Training Service Service Status> - (Output) Status of a TrainingService. Structure is documented below.
- update
Time String - (Output) Time when the TrainingService was most recently updated.
- create
Time string - (Output) Time when the TrainingService was created.
- deployment
Specs AiModel Development Cluster Runtime Training Service Deployment Spec[] - (Output) Orchestrator-agnostic deployment spec. Structure is documented below.
- gcs
Bucket string - (Output) Cloud Storage bucket name dynamically GCSFuse-mounted on all nodes, for storing checkpoints and other service outputs.
- image
Uri string - (Output) Cloud Storage URI of the squashfs (.sqsh) container image.
- name string
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - service
Statuses AiModel Development Cluster Runtime Training Service Service Status[] - (Output) Status of a TrainingService. Structure is documented below.
- update
Time string - (Output) Time when the TrainingService was most recently updated.
- create_
time str - (Output) Time when the TrainingService was created.
- deployment_
specs Sequence[AiModel Development Cluster Runtime Training Service Deployment Spec] - (Output) Orchestrator-agnostic deployment spec. Structure is documented below.
- gcs_
bucket str - (Output) Cloud Storage bucket name dynamically GCSFuse-mounted on all nodes, for storing checkpoints and other service outputs.
- image_
uri str - (Output) Cloud Storage URI of the squashfs (.sqsh) container image.
- name str
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - service_
statuses Sequence[AiModel Development Cluster Runtime Training Service Service Status] - (Output) Status of a TrainingService. Structure is documented below.
- update_
time str - (Output) Time when the TrainingService was most recently updated.
- create
Time String - (Output) Time when the TrainingService was created.
- deployment
Specs List<Property Map> - (Output) Orchestrator-agnostic deployment spec. Structure is documented below.
- gcs
Bucket String - (Output) Cloud Storage bucket name dynamically GCSFuse-mounted on all nodes, for storing checkpoints and other service outputs.
- image
Uri String - (Output) Cloud Storage URI of the squashfs (.sqsh) container image.
- name String
- (Output)
Identifier. Resource name of the TrainingService. Format:
projects/{project}/locations/{location}/modelDevelopmentClusters/{cluster}/trainingServices/{training_service} - service
Statuses List<Property Map> - (Output) Status of a TrainingService. Structure is documented below.
- update
Time String - (Output) Time when the TrainingService was most recently updated.
AiModelDevelopmentClusterRuntimeTrainingServiceDeploymentSpec, AiModelDevelopmentClusterRuntimeTrainingServiceDeploymentSpecArgs
- Worker
Node intCount - (Output) Number of worker nodes to allocate.
- Worker
Node stringPool Id - (Output)
ID of a node pool on the parent ModelDevelopmentCluster (see
ModelDevelopmentCluster.node_pools[].id) on which to place the worker nodes.
- Worker
Node intCount - (Output) Number of worker nodes to allocate.
- Worker
Node stringPool Id - (Output)
ID of a node pool on the parent ModelDevelopmentCluster (see
ModelDevelopmentCluster.node_pools[].id) on which to place the worker nodes.
- worker_
node_ numbercount - (Output) Number of worker nodes to allocate.
- worker_
node_ stringpool_ id - (Output)
ID of a node pool on the parent ModelDevelopmentCluster (see
ModelDevelopmentCluster.node_pools[].id) on which to place the worker nodes.
- worker
Node IntegerCount - (Output) Number of worker nodes to allocate.
- worker
Node StringPool Id - (Output)
ID of a node pool on the parent ModelDevelopmentCluster (see
ModelDevelopmentCluster.node_pools[].id) on which to place the worker nodes.
- worker
Node numberCount - (Output) Number of worker nodes to allocate.
- worker
Node stringPool Id - (Output)
ID of a node pool on the parent ModelDevelopmentCluster (see
ModelDevelopmentCluster.node_pools[].id) on which to place the worker nodes.
- worker_
node_ intcount - (Output) Number of worker nodes to allocate.
- worker_
node_ strpool_ id - (Output)
ID of a node pool on the parent ModelDevelopmentCluster (see
ModelDevelopmentCluster.node_pools[].id) on which to place the worker nodes.
- worker
Node NumberCount - (Output) Number of worker nodes to allocate.
- worker
Node StringPool Id - (Output)
ID of a node pool on the parent ModelDevelopmentCluster (see
ModelDevelopmentCluster.node_pools[].id) on which to place the worker nodes.
AiModelDevelopmentClusterRuntimeTrainingServiceServiceStatus, AiModelDevelopmentClusterRuntimeTrainingServiceServiceStatusArgs
AiModelDevelopmentClusterRuntimeXmanager, AiModelDevelopmentClusterRuntimeXmanagerArgs
- Api
Endpoint string - (Output) The gRPC API endpoint (Experiment State Server) of the XManager instance. Empty unless the instance is running.
- Errors
List<Ai
Model Development Cluster Runtime Xmanager Error> - (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - Instance
Id string - (Output) The XManager instance ID (for example, "xmc-").
- State string
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
- Ui
Endpoint string - (Output) The UI endpoint of the XManager instance. Empty unless the instance is running.
- Api
Endpoint string - (Output) The gRPC API endpoint (Experiment State Server) of the XManager instance. Empty unless the instance is running.
- Errors
[]Ai
Model Development Cluster Runtime Xmanager Error - (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - Instance
Id string - (Output) The XManager instance ID (for example, "xmc-").
- State string
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
- Ui
Endpoint string - (Output) The UI endpoint of the XManager instance. Empty unless the instance is running.
- api_
endpoint string - (Output) The gRPC API endpoint (Experiment State Server) of the XManager instance. Empty unless the instance is running.
- errors list(object)
- (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - instance_
id string - (Output) The XManager instance ID (for example, "xmc-").
- state string
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
- ui_
endpoint string - (Output) The UI endpoint of the XManager instance. Empty unless the instance is running.
- api
Endpoint String - (Output) The gRPC API endpoint (Experiment State Server) of the XManager instance. Empty unless the instance is running.
- errors
List<Ai
Model Development Cluster Runtime Xmanager Error> - (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - instance
Id String - (Output) The XManager instance ID (for example, "xmc-").
- state String
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
- ui
Endpoint String - (Output) The UI endpoint of the XManager instance. Empty unless the instance is running.
- api
Endpoint string - (Output) The gRPC API endpoint (Experiment State Server) of the XManager instance. Empty unless the instance is running.
- errors
Ai
Model Development Cluster Runtime Xmanager Error[] - (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - instance
Id string - (Output) The XManager instance ID (for example, "xmc-").
- state string
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
- ui
Endpoint string - (Output) The UI endpoint of the XManager instance. Empty unless the instance is running.
- api_
endpoint str - (Output) The gRPC API endpoint (Experiment State Server) of the XManager instance. Empty unless the instance is running.
- errors
Sequence[Ai
Model Development Cluster Runtime Xmanager Error] - (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - instance_
id str - (Output) The XManager instance ID (for example, "xmc-").
- state str
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
- ui_
endpoint str - (Output) The UI endpoint of the XManager instance. Empty unless the instance is running.
- api
Endpoint String - (Output) The gRPC API endpoint (Experiment State Server) of the XManager instance. Empty unless the instance is running.
- errors List<Property Map>
- (Output)
The
Statustype defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by gRPC. EachStatusmessage contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the API Design Guide. Structure is documented below. - instance
Id String - (Output) The XManager instance ID (for example, "xmc-").
- state String
- (Output) The current state of the XManager instance. Possible values: PROVISIONING RUNNING DELETING ERROR
- ui
Endpoint String - (Output) The UI endpoint of the XManager instance. Empty unless the instance is running.
AiModelDevelopmentClusterRuntimeXmanagerError, AiModelDevelopmentClusterRuntimeXmanagerErrorArgs
- Code int
- (Output) The status code, which should be an enum value of google.rpc.Code.
- Details
List<Immutable
Dictionary<string, string>> - (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- Message string
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
- Code int
- (Output) The status code, which should be an enum value of google.rpc.Code.
- Details []map[string]string
- (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- Message string
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
- code number
- (Output) The status code, which should be an enum value of google.rpc.Code.
- details list(map(string))
- (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- message string
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
- code Integer
- (Output) The status code, which should be an enum value of google.rpc.Code.
- details List<Map<String,String>>
- (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- message String
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
- code number
- (Output) The status code, which should be an enum value of google.rpc.Code.
- details {[key: string]: string}[]
- (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- message string
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
- code int
- (Output) The status code, which should be an enum value of google.rpc.Code.
- details Sequence[Mapping[str, str]]
- (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- message str
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
- code Number
- (Output) The status code, which should be an enum value of google.rpc.Code.
- details List<Map<String>>
- (Output) A list of messages that carry the error details. There is a common set of message types for APIs to use.
- message String
- (Output) A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the google.rpc.Status.details field, or localized by the client.
Import
ModelDevelopmentCluster can be imported using any of these accepted formats:
projects/{{project}}/locations/{{region}}/modelDevelopmentClusters/{{model_development_cluster_id}}{{project}}/{{region}}/{{model_development_cluster_id}}{{region}}/{{model_development_cluster_id}}{{model_development_cluster_id}}
When using the pulumi import command, ModelDevelopmentCluster can be imported using one of the formats above. For example:
$ pulumi import gcp:vertex/aiModelDevelopmentCluster:AiModelDevelopmentCluster default projects/{{project}}/locations/{{region}}/modelDevelopmentClusters/{{model_development_cluster_id}}
$ pulumi import gcp:vertex/aiModelDevelopmentCluster:AiModelDevelopmentCluster default {{project}}/{{region}}/{{model_development_cluster_id}}
$ pulumi import gcp:vertex/aiModelDevelopmentCluster:AiModelDevelopmentCluster default {{region}}/{{model_development_cluster_id}}
$ pulumi import gcp:vertex/aiModelDevelopmentCluster:AiModelDevelopmentCluster default {{model_development_cluster_id}}
To learn more about importing existing cloud resources, see Importing resources.
Package Details
- Repository
- Google Cloud (GCP) pulumi/pulumi-gcp
- License
- Apache-2.0
- Notes
- This Pulumi package is based on the
google-betaTerraform Provider.
published on Wednesday, Nov 18, 2026 by Pulumi