gcloud cluster-director clusters create

gcloud cluster-director clusters create - creates a Cluster Director resource

gcloud cluster-director clusters create (CLUSTER : --location=LOCATION) (--buckets=[id=ID],[name=NAME] --create-buckets=[enableAutoclass=ENABLEAUTOCLASS],[enableHNS=ENABLEHNS],[id=ID],[name=NAME],[storageClass=STORAGECLASS] --create-filestores=[capacityGb=CAPACITYGB],[description=DESCRIPTION],[fileshare=FILESHARE],[id=ID],[name=NAME],[protocol=PROTOCOL],[tier=TIER] --create-lustres=[capacityGb=CAPACITYGB],[description=DESCRIPTION],[filesystem=FILESYSTEM],[id=ID],[name=NAME],[perUnitStorageThroughput=PERUNITSTORAGETHROUGHPUT] --description=DESCRIPTION --filestores=[id=ID],[name=NAME] --flex-start-instances=[id=ID],[machineType=MACHINETYPE],[maxDuration=MAXDURATION],[zone=ZONE] --labels=[LABELS,…] --lustres=[id=ID],[name=NAME] --on-demand-instances=[id=ID],[machineType=MACHINETYPE],[zone=ZONE] --reference-architecture=REFERENCE_ARCHITECTURE --reserved-instances=[id=ID],[reservation=RESERVATION] --spot-instances=[id=ID],[machineType=MACHINETYPE],[terminationAction=TERMINATIONACTION],[zone=ZONE] --create-network=[description=DESCRIPTION],[name=NAME]     | [--network=NETWORK --subnet=SUBNET : --network-project=NETWORK_PROJECT] --slurm-default-partition=SLURM_DEFAULT_PARTITION --slurm-epilog-scripts=[SLURM_EPILOG_SCRIPTS,…] --slurm-login-node=[bootDisk=BOOTDISK],[count=COUNT],[enableOsLogin=ENABLEOSLOGIN],[enablePublicIps=ENABLEPUBLICIPS],[labels=LABELS],[machineType=MACHINETYPE],[startupScript=STARTUPSCRIPT],[storageConfigs=STORAGECONFIGS],[zone=ZONE] --slurm-node-sets=[computeId=COMPUTEID],[computeInstance=COMPUTEINSTANCE],[id=ID],[maxDynamicNodeCount=MAXDYNAMICNODECOUNT],[staticNodeCount=STATICNODECOUNT],[storageConfigs=STORAGECONFIGS] --slurm-partitions=[id=ID],[nodeSetIds=NODESETIDS] --slurm-prolog-scripts=[SLURM_PROLOG_SCRIPTS,…]) [--async] [GCLOUD_WIDE_FLAG …]

Use one of the following options to create a cluster:

• [Preferred] Use granular flags to define cluster specs.
• Use --reference-architecture to apply pre-configured templates.

quickstart (applied when using --reference-architecture=quickstart):

• Compute: 2x a3-megagpu-8g nodes (Flex Start, max duration 7 days).
• Login Node: 1x n2-standard-16 node.
• Storage: 36TB Lustre storage (scratch disk).

a3-ultra (applied when using --reference-architecture=a3-ultra):

• Compute: 4x reserved nodes (requires specifying --reserved-instances).
• Login Node: 1x n2-standard-16 node.
• Storage: 36TB Lustre storage (scratch disk) and 5.1TB Filestore storage.

a4-high-flex-start (applied when using --reference-architecture=a4-high-flex-start):

• Compute: 4x a4-highgpu-8g nodes (Flex Start, max duration 7 days).
• Login Node: 1x n2-standard-16 node.
• Storage: 18TB Lustre storage (scratch disk) and 2TB Filestore storage.

a4x-high (applied when using --reference-architecture=a4x-high):

• Compute: 18x reserved nodes (requires specifying --reserved-instances).
• Login Node: 1x n2-standard-16 node.
• Storage: 36TB Lustre storage (scratch disk).

g4-flex-start (applied when using --reference-architecture=g4-flex-start):

• Compute: 4x g4-standard-384 nodes (Flex Start, max duration 7 days).
• Login Node: 1x n2-standard-16 node.
• Storage: 36TB Lustre storage (scratch disk) and 10.2TB Filestore storage.

h4d-highmem-flex-start (applied when using --reference-architecture=h4d-highmem-flex-start):

• Compute: 4x h4d-highmem-192 nodes (Flex Start, max duration 7 days).
• Login Node: 1x n2-standard-16 node.
• Storage: 18TB Lustre storage (scratch disk) and 1TB Filestore storage.

Please refer to the examples below for more details.

gcloud cluster-director clusters create my-cluster --location=us-central1 --description="My cluster description" --labels=env=prod,client=gcloud-cli --create-network=name=network0 --create-filestores=name=locations/us-central1-a/instances/filestore0,tier=ZONAL,capacityGb={filestoreSize},fileshare={fileshare} --filestores=locations/us-central1-a/instances/filestore1 --create-buckets=name=bucket0 --buckets=bucket1 --create-lustres=name=locations/us-central1-a/instances/lustre0,capacityGb={lustreSize},filesystem={filesystem} --lustres=locations/us-central1-a/instances/lustre1 --reserved-instances=id=compute0,reservation=zones/us-central1-a/reservations/{reservation} --slurm-login-node=machineType={machineType},zone=us-central1-a --slurm-node-sets=id=nodeset0,computeId=compute0 --slurm-partitions=id=partition0,nodesetIds=[nodeset0] --slurm-default-partition=partition0 --format=json

To create a cluster my-cluster in location us-central1 using the quickstart setup with default cluster parameters, run the following example:

gcloud cluster-director clusters create my-cluster --location=us-central1 --reference-architecture=quickstart --create-network=name=network0

To create a cluster my-cluster in location us-central1 using a pre-defined reference architecture choice (such as g4-flex-start), run the following example:

gcloud cluster-director clusters create my-cluster --location=us-central1 --reference-architecture=g4-flex-start --create-network=name=network0

Cluster resource - Name of the cluster resource. Formats: cluster | projects/{project}/locations/{locations}/clusters/{cluster} The arguments in this group can be used to specify the attributes of this resource. (NOTE) Some attributes are not given arguments in this group but can be set in other ways.

To set the project attribute:

• provide the argument cluster on the command line with a fully specified name;
• provide the argument --project on the command line;
• set the property core/project.

This must be specified.

CLUSTER

ID of the cluster or fully qualified identifier for the cluster.

To set the cluster attribute:

• provide the argument cluster on the command line.

This positional argument must be specified if any of the other arguments in this group are specified.

--location=LOCATION

Location ID of the cluster resource.

To set the location attribute:

• provide the argument cluster on the command line with a fully specified name;
• provide the argument --location on the command line.

Flag Configurations to define cluster spec.

At least one of these must be specified:

--buckets=[id=ID],[name=NAME]

Reference of existing Google Cloud Storage bucket.

id: ID of the bucket resource, used to refer to this resource in storage-configs. name: Name of the existing Cloud Storage bucket to import from your project in the format of bucket-name.

For example --buckets id=my-bucket,name=bucket-1.

Sets buckets value. Shorthand Example:

--buckets=id=string,name=string --buckets=id=string,name=string

JSON Example:

--buckets='[{"id": "string", "name": "string"}]'

File Example:

--buckets=path_to_file.(yaml|json)

--create-buckets=[enableAutoclass=ENABLEAUTOCLASS],[enableHNS=ENABLEHNS],[id=ID],[name=NAME],[storageClass=STORAGECLASS]

Parameters to create a Google Cloud Storage bucket.

Required fields:

• id
• name

For example --create-buckets id=my-bucket,name=bucket-1,storageClass=STANDARD,enableHNS=true

Supported storageClass values:

• STANDARD
• NEARLINE
• COLDLINE
• ARCHIVE

Defaults:

• storageClass: STANDARD

Note:

• Either storageClass or enableAutoclass can be set.
• HNS: Hierarchical namespace. Hierarchical namespace buckets cannot use Autoclass.

Sets create_buckets value. Shorthand Example:

--create-buckets=enableAutoclass=boolean,enableHNS=boolean,id=string,name=string,storageClass=string --create-buckets=enableAutoclass=boolean,enableHNS=boolean,id=string,name=string,storageClass=string

JSON Example:

--create-buckets='[{"enableAutoclass": boolean, "enableHNS": boolean, "id": "string", "name": "string", "storageClass": "string"}]'

File Example:

--create-buckets=path_to_file.(yaml|json)

--create-filestores=[capacityGb=CAPACITYGB],[description=DESCRIPTION],[fileshare=FILESHARE],[id=ID],[name=NAME],[protocol=PROTOCOL],[tier=TIER]

Parameters to create a filestore instance.

Required fields:

• id
• name
• tier
• capacityGb
• fileshare

For example --create-filestores id=my-fs,name=locations/us-central1/instances/filestore-1,tier=REGIONAL,capacityGb=1024,fileshare=share1

id: Alphanumeric identifier of the filestore resource. Can be used to refer to this resource in storage-configs. name: Name of the filestore instance to create in your project in the format of locations/us-central1/instances/filestore-1.

capacityGb: Size of the filestore in GiB. Must be between 1024 and 102400, and must meet scalability requirements described at https://cloud.google.com/filestore/docs/service-tiers.

fileshare: The directory on a Filestore instance where all shared files are stored. Must match the regex [a-z]([-a-z0-9]*[a-z0-9])?, be 1-63 characters in length, and comply with RFC1035.

Supported protocol values:

• NFSV3
• NFSV41
• If not specified, defaults to NFSV3

Defaults:

• protocol: NFSV3.

Sets create_filestores value. Shorthand Example:

--create-filestores=capacityGb=int,description=string,fileshare=string,id=string,name=string,protocol=string,tier=string --create-filestores=capacityGb=int,description=string,fileshare=string,id=string,name=string,protocol=string,tier=string

JSON Example:

--create-filestores='[{"capacityGb": int, "description": "string", "fileshare": "string", "id": "string", "name": "string", "protocol": "string", "tier": "string"}]'

File Example:

--create-filestores=path_to_file.(yaml|json)

--create-lustres=[capacityGb=CAPACITYGB],[description=DESCRIPTION],[filesystem=FILESYSTEM],[id=ID],[name=NAME],[perUnitStorageThroughput=PERUNITSTORAGETHROUGHPUT]

Parameters to create a Lustre instance.

Required fields:

• id
• name
• capacityGb
• filesystem.

id: ID of the lustre resource, used to refer to this resource in storage-configs.

name: Name of the Managed Lustre instance to create in your project in the format of locations/us-central1/instances/lustre-1.

For example --create-lustres id=my-lustre,name=locations/us-central1/instances/lustre-1,capacityGb=1024,filesystem=fs-1,perUnitStorageThroughput=1000

Values for perUnitStorageThroughput: 125, 250, 500, 1000.

Sets create_lustres value. Shorthand Example:

--create-lustres=capacityGb=int,description=string,filesystem=string,id=string,name=string,perUnitStorageThroughput=int --create-lustres=capacityGb=int,description=string,filesystem=string,id=string,name=string,perUnitStorageThroughput=int

JSON Example:

--create-lustres='[{"capacityGb": int, "description": "string", "filesystem": "string", "id": "string", "name": "string", "perUnitStorageThroughput": int}]'

File Example:

--create-lustres=path_to_file.(yaml|json)

--description=DESCRIPTION

Description of the cluster.

For example --description "My cluster"

--filestores=[id=ID],[name=NAME]

Reference of existing filestore instance.

id: ID of the filestore resource, used to refer to this resource in storage-configs. name: Name of the existing filestore instance to import from your project in the format of locations/us-central1/instances/filestore-1.

For example --filestores id=my-fs,name=locations/us-central1/instances/filestore-1.

Sets filestores value. Shorthand Example:

--filestores=id=string,name=string --filestores=id=string,name=string

JSON Example:

--filestores='[{"id": "string", "name": "string"}]'

File Example:

--filestores=path_to_file.(yaml|json)

--flex-start-instances=[id=ID],[machineType=MACHINETYPE],[maxDuration=MAXDURATION],[zone=ZONE]

Parameters to define cluster Flex Start instances.

For example --flex-start-instances id=c1,zone=us-central1-a,machineType=n1-standard-1,maxDuration=10000s.

Sets flex_start_instances value. Shorthand Example:

--flex-start-instances=id=string,machineType=string,maxDuration=string,zone=string --flex-start-instances=id=string,machineType=string,maxDuration=string,zone=string

JSON Example:

--flex-start-instances='[{"id": "string", "machineType": "string", "maxDuration": "string", "zone": "string"}]'

File Example:

--flex-start-instances=path_to_file.(yaml|json)

--labels=[LABELS,…]

Cluster labels as key value pairs.

For example --labels key1=value1,key2=value2.

Sets labels value. Shorthand Example:

--labels={string=string}

JSON Example:

--labels='{"string": "string"}'

File Example:

--labels=path_to_file.(yaml|json)

--lustres=[id=ID],[name=NAME]

Reference of existing Lustre instance.

id: ID of the lustre resource, used to refer to this resource in storage-configs. name: Name of the existing Managed Lustre instance to import from your project in the format of locations/us-central1/instances/lustre-1.

For example --lustres id=my-lustre,name=locations/us-central1/instances/lustre-1.

Sets lustres value. Shorthand Example:

--lustres=id=string,name=string --lustres=id=string,name=string

JSON Example:

--lustres='[{"id": "string", "name": "string"}]'

File Example:

--lustres=path_to_file.(yaml|json)

--on-demand-instances=[id=ID],[machineType=MACHINETYPE],[zone=ZONE]

Parameters to define cluster on demand instances.

For example --on-demand-instances id=c1,zone=us-central1-a,machineType=n1-standard-1.

Sets on_demand_instances value. Shorthand Example:

--on-demand-instances=id=string,machineType=string,zone=string --on-demand-instances=id=string,machineType=string,zone=string

JSON Example:

--on-demand-instances='[{"id": "string", "machineType": "string", "zone": "string"}]'

File Example:

--on-demand-instances=path_to_file.(yaml|json)

--reference-architecture=REFERENCE_ARCHITECTURE

Reference a pre-defined architecture. Available options:

• a3-ultra: 4x reserved nodes (requires --reserved-instances), 1x login node (n2-standard-16), 36TB Lustre storage, and 5.1TB Filestore storage.
• a4-high-flex-start: 4x a4-highgpu-8g compute nodes (Flex Start), 1x login node (n2-standard-16), 18TB Lustre storage, and 2TB Filestore storage.
• a4x-high: 18x reserved nodes (requires --reserved-instances), 1x login node (n2-standard-16), and 36TB Lustre storage.
• g4-flex-start: 4x g4-standard-384 compute nodes (Flex Start), 1x login node (n2-standard-16), 36TB Lustre storage, and 10.2TB Filestore storage.
• h4d-highmem-flex-start: 4x h4d-highmem-192 compute nodes (Flex Start), 1x login node (n2-standard-16), 18TB Lustre storage, and 1TB Filestore storage.
• quickstart: 2x a3-megagpu-8g compute nodes (Flex Start, max duration 7 days), 1x login node (n2-standard-16), and 36TB Lustre storage. REFERENCE_ARCHITECTURE must be one of: a3-ultra, a4-high-flex-start, a4x-high, g4-flex-start, h4d-highmem-flex-start, quickstart.

--reserved-instances=[id=ID],[reservation=RESERVATION]

Parameters to define cluster reserved instances.

For example --reserved-instances id=c1,reservation=zones/us-central1-a/reservations/reservation-1.

Sets reserved_instances value. Shorthand Example:

--reserved-instances=id=string,reservation=string --reserved-instances=id=string,reservation=string

JSON Example:

--reserved-instances='[{"id": "string", "reservation": "string"}]'

File Example:

--reserved-instances=path_to_file.(yaml|json)

--spot-instances=[id=ID],[machineType=MACHINETYPE],[terminationAction=TERMINATIONACTION],[zone=ZONE]

Parameters to define cluster spot instances.

For example --spot-instances id=c1,zone=us-central1-a,machineType=n1-standard-1.

Sets spot_instances value. Shorthand Example:

--spot-instances=id=string,machineType=string,terminationAction=string,zone=string --spot-instances=id=string,machineType=string,terminationAction=string,zone=string

JSON Example:

--spot-instances='[{"id": "string", "machineType": "string", "terminationAction": "string", "zone": "string"}]'

File Example:

--spot-instances=path_to_file.(yaml|json)

Network configuration for the cluster.

At most one of these can be specified:

--create-network=[description=DESCRIPTION],[name=NAME]

Parameters to create a VPC (Virtual Private Cloud) network.

Required fields:

• name Name: Must match the regex [a-z]([-a-z0-9]*[a-z0-9])?, be 1-63 characters in length, and comply with RFC1035.

Description: A description of the network. Maximum of 2048 characters.

For example --create-network name=network-1,description="My network".

Sets create_network value. Shorthand Example:

--create-network=description=string,name=string

JSON Example:

--create-network='{"description": "string", "name": "string"}'

File Example:

--create-network=path_to_file.(yaml|json)

Or at least one of these can be specified:

Use an existing network source for the cluster.

--network=NETWORK

Reference of existing network name. If the network is in a different project (Shared VPC), specify the project ID using --network-project.

For example --network network-1

This flag argument must be specified if any of the other arguments in this group are specified.

--subnet=SUBNET

Reference of existing subnetwork name. If the subnetwork is in a different project (Shared VPC), specify the project ID using --network-project.

For example --subnet regions/us-central1/subnetworks/subnet-1

This flag argument must be specified if any of the other arguments in this group are specified.

--network-project=NETWORK_PROJECT

Project ID of the project containing the network and subnetwork resources, if different from the cluster project (e.g. for Shared VPC).

Orchestrator configuration for the cluster.

Slurm orchestrator configuration for the cluster.

--slurm-default-partition=SLURM_DEFAULT_PARTITION

Parameters to define slurm cluster default partition.

For example --slurm-default-partition p1

--slurm-epilog-scripts=[SLURM_EPILOG_SCRIPTS,…]

Slurm epilog bash scripts.

For example --slurm-epilog-scripts script1.sh,script2.sh

--slurm-login-node=[bootDisk=BOOTDISK],[count=COUNT],[enableOsLogin=ENABLEOSLOGIN],[enablePublicIps=ENABLEPUBLICIPS],[labels=LABELS],[machineType=MACHINETYPE],[startupScript=STARTUPSCRIPT],[storageConfigs=STORAGECONFIGS],[zone=ZONE]

Parameters to define slurm cluster login node.

Required:

• machineType
• zone

For example --slurm-login-node machineType=n1-standard-1,zone=us-central1-a,count=1

Defaults:

• count: 1
• enableOsLogin: true
• enablePublicIPs: true
• bootDisk.sizeGb: 100

Note:

• startupScript:
  • Either str or file_path
  • For file_path, only bash file format (.sh or .bash) is supported.
  • For file_path, only absolute path is supported.
• bootDisk:
  • Must be 50 GB or greater.

  Sets slurm_login_node value.

Shorthand Example:

--slurm-login-node=bootDisk={sizeGb=int,type=string},count=int,enableOsLogin=boolean,enablePublicIps=boolean,labels={string=string},machineType=string,startupScript=string,storageConfigs=[{id=string,localMount=string}],zone=string

JSON Example:

--slurm-login-node='{"bootDisk": {"sizeGb": int, "type": "string"}, "count": int, "enableOsLogin": boolean, "enablePublicIps": boolean, "labels": {"string": "string"}, "machineType": "string", "startupScript": "string", "storageConfigs": [{"id": "string", "localMount": "string"}], "zone": "string"}'

File Example:

--slurm-login-node=path_to_file.(yaml|json)

--slurm-node-sets=[computeId=COMPUTEID],[computeInstance=COMPUTEINSTANCE],[id=ID],[maxDynamicNodeCount=MAXDYNAMICNODECOUNT],[staticNodeCount=STATICNODECOUNT],[storageConfigs=STORAGECONFIGS]

Parameters to define slurm cluster nodeset config.

Required fields:

• id

For example --slurm-node-sets id=ns1,computeId=c1,staticNodeCount=1,maxDynamicNodeCount=2,computeInstance=[startupScript="echo hello",labels="k1=v1"]

Defaults:

Note:

• startupScript:
  • Either str or file_path
  • For file_path, only bash file format (.sh or .bash) is supported.
  • For file_path, only absolute path is supported.

  Sets slurm_node_sets value.

Shorthand Example:

--slurm-node-sets=computeId=string,computeInstance={bootDisk={sizeGb=int,type=string},labels={string=string},startupScript=string},id=string,maxDynamicNodeCount=int,staticNodeCount=int,storageConfigs=[{id=string,localMount=string}] --slurm-node-sets=computeId=string,computeInstance={bootDisk={sizeGb=int,type=string},labels={string=string},startupScript=string},id=string,maxDynamicNodeCount=int,staticNodeCount=int,storageConfigs=[{id=string,localMount=string}]

JSON Example:

--slurm-node-sets='[{"computeId": "string", "computeInstance": {"bootDisk": {"sizeGb": int, "type": "string"}, "labels": {"string": "string"}, "startupScript": "string"}, "id": "string", "maxDynamicNodeCount": int, "staticNodeCount": int, "storageConfigs": [{"id": "string", "localMount": "string"}]}]'

File Example:

--slurm-node-sets=path_to_file.(yaml|json)

--slurm-partitions=[id=ID],[nodeSetIds=NODESETIDS]

Parameters to define slurm cluster partitions.

For example --slurm-partitions id=p1,nodesetIds=[ns1,ns2].

Sets slurm_partitions value. Shorthand Example:

--slurm-partitions=id=string,nodeSetIds=[string] --slurm-partitions=id=string,nodeSetIds=[string]

JSON Example:

--slurm-partitions='[{"id": "string", "nodeSetIds": ["string"]}]'

File Example:

--slurm-partitions=path_to_file.(yaml|json)

--slurm-prolog-scripts=[SLURM_PROLOG_SCRIPTS,…]

Slurm prolog bash scripts.

For example --slurm-prolog-scripts script1.sh,script2.sh

--async

Return immediately, without waiting for the operation in progress to complete.

Run $ gcloud help for details.