Create an AI-optimized MIG with A4 or A3 Ultra machine type (original) (raw)

Technology areas
- Overview
- Guides
- Resources
Cross-product tools
Console
Discover
Overview
Choose a consumption option
Terminology
Get started
Cluster creation overview
Recommended configurations
Deploy infrastructure
Deployment options overview
Compact placement policy and workload policy overview
Deploy AI-optimized VMs and clusters
- Create GKE clusters
  * Create an AI-optimized GKE cluster with default configuration
  * Create a custom AI-optimized GKE cluster which uses A4X Max
  * Create a custom AI-optimized GKE cluster which uses A4X
  * Create a custom AI-optimized GKE cluster which uses A4 or A3 Ultra
  * Create GKE Standard clusters which use A3 Mega or A3 High
  * Create GKE Autopilot clusters which use A3 Mega or A3 High
Run workloads
AI workload tutorials
- Overview
- TPU
  * Serve Qwen2-7B with vLLM on TPUs
  * Serve Qwen2-7B-Instruct with vLLM on TPUs
  * Serve Qwen3-8B-Base with vLLM on TPUs
  * Serve Llama-3.1-8B with vLLM on TPUs
Manage infrastructure
Manage GKE clusters
Test and optimize
Optimize cluster networking by using NCCL/gIB
Run NCCL tests
- Run NCCL on Compute Engine instances
- Run NCCL on GKE clusters that use default configuration
- Run NCCL on custom GKE clusters that use A4X Max
- Run NCCL on custom GKE clusters that use A4X
- Run NCCL on custom GKE clusters that use A4 or A3 Ultra
- Run NCCL on custom GKE clusters that use A3 Mega or A3 High
- Run NCCL on Slurm clusters
Collect and understand NCCL logs for troubleshooting
Test clusters
Monitor
Monitor VMs and Slurm clusters
Troubleshoot
Known issues
Troubleshoot slow performance
Troubleshoot reporting a faulty host
Troubleshoot CoMMA

Create an AI-optimized MIG with A4 or A3 Ultra machine type

This document describes how to create a managed instance group (MIG) that uses A4 or A3 Ultra machine types. To learn more about these accelerator-optimized machine types, see A4 orA3 Ultra.

Creating a MIG lets you manage multiple virtual machines (VMs) as a single entity. Each VM in a MIG is based on an instance template. By automatically managing the VMs in the group, MIGs offer high availability and scalability. To learn more about MIGs, seeManaged instance groupsin the Compute Engine documentation.

To learn about other ways to create VMs or clusters, seeVM and cluster creation overviewpage.

Limitations

When you create a MIG with A4 or A3 Ultra VMs, consider the limitations for both the machine type and the MIG.

Limitations for A4 or A3 Ultra VMs

Depending on the machine series that the VMs in your MIG use, the following limitations apply:

A4

You don't receive sustained use discounts andflexible committed usediscounts for instances that use an A4 machine type.
You can only use an A4 machine type in certain regions and zones.
You can't use Persistent Disk (regional or zonal). You can only useGoogle Cloud Hyperdisk.
The A4 machine type is only available on the Emerald Rapids CPU platform.
You can't change the machine type of an instance to or from A4 machine type. You must create a new instance with this machine type.
A4 machine types don't supportsole-tenancy.
You can't run Windows operating systems on an A4 machine type.
For A4 instances, when you use ethtool -S to monitor GPU networking, physical port counters that end in _phy don't update. This is expected behavior for instances that use the MRDMA Virtual Function (VF) architecture. For more information, seeMRDMA functions and network monitoring tools.
You can't attach Hyperdisk ML disks that were created before February 4, 2026 to A4 machine types.

A3 Ultra

You don't receive sustained use discounts andflexible committed usediscounts for instances that use an A3 Ultra machine type.
You can only use an A3 Ultra machine type in certain regions and zones.
You can't use Persistent Disk (regional or zonal). You can only useGoogle Cloud Hyperdisk.
The A3 Ultra machine type is only available on the Emerald Rapids CPU platform.
Machine type changes aren't supported for A3 Ultra machine type. To switch to or from this machine type, you must create a new instance.
You can't run Windows operating systems on an A3 Ultra machine type.
A3 Ultra machine types don't supportsole-tenancy.
For A3 Ultra instances, when you use ethtool -S to monitor GPU networking, physical port counters that end in _phy don't update. This is expected behavior for instances that use the MRDMA Virtual Function (VF) architecture. For more information, seeMRDMA functions and network monitoring tools.

Limitations for the MIG

When you create a MIG with A4 or A3 Ultra VMs, the following limitations apply:

You must turn off repairs in the MIG, as described in this document.
If you create a regional MIG that uses RDMA networking, then the MIG can only create A4 or A3 Ultra instances in the zone where youconfigured the network profile for RDMA.
You can't configureinstance flexibilityin the MIG.
If you apply a workload policy to a MIG, you cannot change the policy in the MIG when the group has A4 or A3 Ultra instances in it. To change the policy, you must firstresizethe MIG to zero.
You can't use an instance template that specifies a placement policy to create a MIG that uses a workload policy.

Limitations for MIG resize requests

For MIG resize requests, the following limitations apply:

In regional MIGs, you can use only the ANY_SINGLE_ZONE target distribution shape(Preview).
You can only set thestandby pool mode of the MIGto manual (default).
You can't set autoscaling.
If the MIG contains accepted resize requests, then you can't change thetarget size of the MIG.
You can't delete or abandon the managed instances in a CREATING status that the MIG creates for a resize request. To delete those managed instances, you mustcancel the resize request.

Before you begin

Before creating a MIG, if you haven't already done so, complete the following steps:

Choose a consumption option: your choice of consumption option determines how you get and use GPU resources. To learn more, seeChoose a consumption option.
Obtain capacity: the process to obtain capacity differs for each consumption option. To learn about the process to obtain capacity for your chosen consumption option, seeCapacity overview.

Required roles

To get the permissions that you need to create a MIG, ask your administrator to grant you theCompute Instance Admin (v1) (roles/compute.instanceAdmin.v1) IAM role on the project. For more information about granting roles, see Manage access to projects, folders, and organizations.

This predefined role contains the permissions required to create a MIG. To see the exact permissions that are required, expand the Required permissions section:

Required permissions

The following permissions are required to create a MIG:

To create a MIG:compute.instanceGroupManagers.create on the project

You might also be able to get these permissions with custom roles or other predefined roles.

Overview

Creating a MIG with the A4 or A3 Ultra machine type includes the following steps:

Create VPC networks
Optional: Create a workload policy
Create an instance template
Create a MIG

Create VPC networks

To set up the network for A4 or A3 Ultra machine types, create three VPC networks for the following network interfaces:

2 regular VPC networks for the gVNIC network interfaces (NICs). These are used for host to host communication.
1 VPC network with the RoCE network profile for the CX-7 NICs. The RoCE VPC network needs to have 8 subnets, one subnet for each CX-7 NIC. These NICs use RDMA over Converged Ethernet (RoCE), providing the high-bandwidth, low-latency communication that's essential for GPU to GPU communication.

For more information about NIC arrangement, seeReview network bandwidth and NIC arrangement.

Create the networks either manually by following the instruction guides or automatically by using the provided script.

Instruction guides

To create the networks, you can use the following instructions:

To create the regular VPC networks for the gVNICs, seeCreate and manage Virtual Private Cloud networks.
To create the RoCE VPC network, seeCreate a Virtual Private Cloud network for RDMA NICs.

For these VPC networks, we recommend setting themaximum transmission unit (MTU) to a larger value. For A4 or A3 Ultra machine types, the recommended MTU is 8896 bytes. To review the recommended MTU settings for other GPU machine types, seeMTU settings for GPU machine types.

Script

To create the networks, follow these steps.

Use the following script to create VPC networks for the gVNICs and CX-7 NICs.

#!/bin/bash

Create regular VPC networks and subnets for the gVNICs
for N in $(seq 0 1); do
gcloud compute networks create GVNIC_NAME_PREFIX-net-$N \
--subnet-mode=custom \
--mtu=8896
gcloud compute networks subnets create GVNIC_NAME_PREFIX-sub-$N \
--network=GVNIC_NAME_PREFIX-net-$N \
--region=REGION \
--range=10.$N.0.0/16
gcloud compute firewall-rules create GVNIC_NAME_PREFIX-internal-$N \
--network=GVNIC_NAME_PREFIX-net-$N \
--action=ALLOW \
--rules=tcp:0-65535,udp:0-65535,icmp \
--source-ranges=10.0.0.0/8
done

Create SSH firewall rules
gcloud compute firewall-rules create GVNIC_NAME_PREFIX-ssh \
--network=GVNIC_NAME_PREFIX-net-0 \
--action=ALLOW \
--rules=tcp:22 \
--source-ranges=IP_RANGE

Assumes that an external IP is only created for vNIC 0
gcloud compute firewall-rules create GVNIC_NAME_PREFIX-allow-ping-net-0 \
--network=GVNIC_NAME_PREFIX-net-0 \
--action=ALLOW \
--rules=icmp \
--source-ranges=IP_RANGE

#!/bin/bash

List and make sure network profiles exist in the machine type's zone
gcloud compute network-profiles list --filter "location.name=ZONE"

Create network for RDMA NICs
gcloud compute networks create RDMA_NAME_PREFIX-mrdma \
--network-profile=ZONE-vpc-roce \
--subnet-mode custom \
--mtu=8896

Create subnets
for N in $(seq 0 7); do
gcloud compute networks subnets create RDMA_NAME_PREFIX-mrdma-sub-$N \
--network=RDMA_NAME_PREFIX-mrdma \
--region=REGION \
--range=10.$((N+2)).0.0/16 # offset to avoid overlap with gVNICs
done

Replace the following:

GVNIC_NAME_PREFIX: the custom name prefix to use for the regular VPC networks and subnets for the gVNICs.
RDMA_NAME_PREFIX: the custom name prefix to use for the RoCE VPC network and subnets for the CX-7 NICs.
ZONE: specify a zone in which the machine type that you want to use is available, such as us-central1-a. For information about regions, seeGPU availability by regions and zones.
REGION: the region where you want to create the subnets. This region must correspond to the zone specified. For example, if your zone is us-central1-a, then your region is us-central1.
IP_RANGE: the IP range to use for theSSH firewall rules.

Optional: To verify that the VPC network resources are created successfully, check the network settings in the Google Cloud console:
1. In the Google Cloud console, go to the VPC networks page.
  Go to VPC networks
2. Search the list for the networks that you created in the previous step.
3. To view the subnets, firewall rules, and other network settings, click the name of the network.

Optional: Create a workload policy

If you want your VMs to be placed in a single or adjacent blocks, then specify VM placement by creating aworkload policy. However, if you want Compute Engine to place your VMs on a specific block, skip this step and provide the name of the block in the reservation affinity when you create the instance template.

To create a workload policy, select one of the following options:

gcloud

To create a workload policy, use the gcloud compute resource-policies create workload-policy command.

For a best-effort placement of A4 VMs, specify the --type=high-throughput flag in the command:
gcloud compute resource-policies create workload-policy WORKLOAD_POLICY_NAME \
--type=high-throughput \
--region=REGION
For strict colocation of A4 or A3 Ultra VMs, specify the--max-topology-distance and --type=high-throughput flags in the command:
gcloud compute resource-policies create workload-policy WORKLOAD_POLICY_NAME \
--type=high-throughput \
--max-topology-distance=TOPOLOGY_DISTANCE \
--region=REGION

Replace the following:

WORKLOAD_POLICY_NAME: the name of the workload policy.
TOPOLOGY_DISTANCE: the maximum topology distance. Specify one of the following values:
- To place A4 or A3 Ultra VMs in the same sub-block: SUBBLOCK
- To place A4 or A3 Ultra VMs in the same block: BLOCK
- To place A4 VMs in the same cluster:CLUSTER
  Note: A shorter maximum distance can reduce the probability of VM availability. For more information, seeMaximum topology distance property.
REGION: the region where you want to create the workload policy. Specify a region in which you want to create the MIG and the machine type that you want to use is available. For information about regions, seeGPU availability by regions and zones.

REST

To create a workload policy, make a POSTrequest to the resourcePolicies.insert method.

For a best-effort placement of A4 VMs, specify the type field in the request as follows:
POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/resourcePolicies
{
"name": "WORKLOAD_POLICY_NAME"
"workloadPolicy": {
"type": "HIGH_THROUGHPUT"
}
}
For strict colocation of A4 or A3 Ultra VMs, specify themaxTopologyDistance and type fields in the request as follows:
POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/resourcePolicies
{
"name": "WORKLOAD_POLICY_NAME"
"workloadPolicy": {
"type": "HIGH_THROUGHPUT",
"maxTopologyDistance": "TOPOLOGY_DISTANCE"
}
}

Replace the following:

PROJECT_ID: your project ID
REGION: the region where you want to create the workload policy. Specify a region in which you want to create the MIG and the machine type that you want to use is available. For information about regions, seeGPU availability by regions and zones.
WORKLOAD_POLICY_NAME: the name of the workload policy.
TOPOLOGY_DISTANCE: the maximum topology distance. Specify one of the following values:
- To place A4 or A3 Ultra VMs in the same sub-block: SUBBLOCK
- To place A4 or A3 Ultra VMs in the same block: BLOCK
- To place A4 VMs in the same cluster: CLUSTER
  Note: A shorter maximum distance can reduce the probability of VM availability. For more information, seeMaximum topology distance property.

Create an instance template

Specify the VM properties for a MIG by creating an instance template.

To create an instance template, select one of the following options:

The following commands also set the access scope for your instances. To simplify permissions management, Google recommends that you set the access scope on an instance tocloud-platform access and then use IAM roles to define what services the instance can access. For more information, seeScopes best practice.

gcloud

The parameters that you need to specify depend on theconsumption option that you are using for this deployment. Select the tab that corresponds to your consumption option's provisioning model.

Flex-start

To create a regional instance template, use the gcloud compute instance-templates create command.

gcloud compute instance-templates create INSTANCE_TEMPLATE_NAME
--machine-type=MACHINE_TYPE
--image-family=IMAGE_FAMILY
--image-project=IMAGE_PROJECT
--instance-template-region=REGION
--boot-disk-type=hyperdisk-balanced
--boot-disk-size=DISK_SIZE
--scopes=cloud-platform
--network-interface=nic-type=GVNIC,network=GVNIC_NAME_PREFIX-net-0,subnet=GVNIC_NAME_PREFIX-sub-0
--network-interface=nic-type=GVNIC,network=GVNIC_NAME_PREFIX-net-1,subnet=GVNIC_NAME_PREFIX-sub-1,no-address --network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-0,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-1,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-2,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-3,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-4,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-5,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-6,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-7,no-address
--reservation-affinity=none
--instance-termination-action=DELETE
--max-run-duration=RUN_DURATION
--maintenance-policy=TERMINATE
--provisioning-model=FLEX_START

Replace the following:

INSTANCE_TEMPLATE_NAME: the name of the instance template.
MACHINE_TYPE: the machine type to use for the VM. Specify either an A4 or A3 Ultra machine type. For more information, seeGPU machine types.
IMAGE_FAMILY: the image family of the OS image that you want to use. For a list of supported operating systems, see Supported operating systems.
IMAGE_PROJECT: the project ID of the OS image.
REGION: the region where you want to create the instance template. Specify a region in which the machine type that you want to use is available. For information about regions, seeGPU availability by regions and zones.
DISK_SIZE: the size of the boot disk in GB.
GVNIC_NAME_PREFIX: the name prefix that you specified when creating the standard VPC networks and subnets that use gVNIC NICs.
RDMA_NAME_PREFIX: the name prefix that you specified when creating the VPC networks and subnets that use RDMA NICs.
RUN_DURATION: the duration you want the requested VMs to run. You must format the value as the number of days, hours, minutes, or seconds followed by d, h, m, and s respectively. For example, specify 30m for 30 minutes or 1d2h3m4s for one day, two hours, three minutes, and four seconds. The value must be between 10 minutes and seven days.

Reservation-bound

To create a regional instance template, use the gcloud compute instance-templates create command.

gcloud compute instance-templates create INSTANCE_TEMPLATE_NAME
--machine-type=MACHINE_TYPE
--image-family=IMAGE_FAMILY
--image-project=IMAGE_PROJECT
--instance-template-region=REGION
--boot-disk-type=hyperdisk-balanced
--boot-disk-size=DISK_SIZE
--scopes=cloud-platform
--network-interface=nic-type=GVNIC,network=GVNIC_NAME_PREFIX-net-0,subnet=GVNIC_NAME_PREFIX-sub-0
--network-interface=nic-type=GVNIC,network=GVNIC_NAME_PREFIX-net-1,subnet=GVNIC_NAME_PREFIX-sub-1,no-address --network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-0,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-1,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-2,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-3,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-4,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-5,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-6,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-7,no-address
--reservation-affinity=specific
--reservation=RESERVATION
--provisioning-model=RESERVATION_BOUND
--instance-termination-action=DELETE
--maintenance-policy=TERMINATE
--restart-on-failure

Replace the following:

INSTANCE_TEMPLATE_NAME: the name of the instance template.
MACHINE_TYPE: the machine type to use for the VM. Specify either an A4 or A3 Ultra machine type. For more information, seeGPU machine types.
IMAGE_FAMILY: the image family of the OS image that you want to use. For a list of supported operating systems, see Supported operating systems.
IMAGE_PROJECT: the project ID of the OS image.
REGION: the region where you want to create the instance template. Specify a region in which the machine type that you want to use is available. For information about regions, seeGPU availability by regions and zones.
DISK_SIZE: the size of the boot disk in GB.
GVNIC_NAME_PREFIX: the name prefix that you specified when creating the standard VPC networks and subnets that use gVNIC NICs.
RDMA_NAME_PREFIX: the name prefix that you specified when creating the VPC networks and subnets that use RDMA NICs.
RESERVATION: either the reservation name or a specific block within a reservation. To get the reservation name or the available blocks, see View reserved capacity. Based on your requirement for instance placement, choose one of the following:
- To create instances across blocks or on a single block:
  projects/RESERVATION_OWNER_PROJECT_ID/reservations/RESERVATION_NAME
  Additionally, for a single block, create the MIG by applying a workload policy that specifies a block collocation (maxTopologyDistance=BLOCK) . Compute Engine then applies the policy to the reservation and creates instances on the same block.
- To create instances on a specific block:
  projects/RESERVATION_OWNER_PROJECT_ID/reservations/RESERVATION_NAME/reservationBlocks/RESERVATION_BLOCK_NAME

Spot

To create a regional instance template, use the gcloud compute instance-templates create command.

gcloud compute instance-templates create INSTANCE_TEMPLATE_NAME
--machine-type=MACHINE_TYPE
--image-family=IMAGE_FAMILY
--image-project=IMAGE_PROJECT
--instance-template-region=REGION
--boot-disk-type=hyperdisk-balanced
--boot-disk-size=DISK_SIZE
--scopes=cloud-platform
--network-interface=nic-type=GVNIC,network=GVNIC_NAME_PREFIX-net-0,subnet=GVNIC_NAME_PREFIX-sub-0
--network-interface=nic-type=GVNIC,network=GVNIC_NAME_PREFIX-net-1,subnet=GVNIC_NAME_PREFIX-sub-1,no-address --network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-0,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-1,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-2,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-3,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-4,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-5,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-6,no-address
--network-interface=nic-type=MRDMA,network=RDMA_NAME_PREFIX-mrdma,subnet=RDMA_NAME_PREFIX-mrdma-sub-7,no-address
--provisioning-model=SPOT
--instance-termination-action=TERMINATION_ACTION
--maintenance-policy=TERMINATE
--no-restart-on-failure

Replace the following:

INSTANCE_TEMPLATE_NAME: the name of the instance template.
MACHINE_TYPE: the machine type to use for the VM. Specify either an A4 or A3 Ultra machine type. For more information, seeGPU machine types.
IMAGE_FAMILY: the image family of the OS image that you want to use. For a list of supported operating systems, see Supported operating systems.
IMAGE_PROJECT: the project ID of the OS image.
REGION: the region where you want to create the instance template. Specify a region in which the machine type that you want to use is available. For information about regions, seeGPU availability by regions and zones.
DISK_SIZE: the size of the boot disk in GB.
GVNIC_NAME_PREFIX: the name prefix that you specified when creating the standard VPC networks and subnets that use gVNIC NICs.
RDMA_NAME_PREFIX: the name prefix that you specified when creating the VPC networks and subnets that use RDMA NICs.
TERMINATION_ACTION: the action to take when Compute Engine preempts the instance, either STOP (default) or DELETE.

REST

The parameters that you need to specify depend on theconsumption option that you are using for this deployment. Select the tab that corresponds to your consumption option's provisioning model.

Flex-start

To create a regional instance template, make a POST request to the regionInstanceTemplates.insert method.

POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/instanceTemplates { "name":"INSTANCE_TEMPLATE_NAME", "properties":{ "machineType":"MACHINE_TYPE", "disks":[ { "boot":true, "initializeParams":{ "diskSizeGb":"DISK_SIZE", "diskType":"hyperdisk-balanced", "sourceImage":"projects/IMAGE_PROJECT/global/images/family/IMAGE_FAMILY" }, "mode":"READ_WRITE", "type":"PERSISTENT" } ], "serviceAccounts": [ { "email": "default", "scopes": [ "https://www.googleapis.com/auth/cloud-platform" ] } ], "networkInterfaces": [ { "accessConfigs": [ { "name": "external-nat", "type": "ONE_TO_ONE_NAT" } ], "network": "projects/NETWORK_PROJECT_ID/global/networks/GVNIC_NAME_PREFIX-net-0", "nicType": "GVNIC", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/GVNIC_NAME_PREFIX-sub-0" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/GVNIC_NAME_PREFIX-net-1", "nicType": "GVNIC", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/GVNIC_NAME_PREFIX-sub-1" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-0" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-1" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-2" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-3" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-4" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-5" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-6" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-7" } ], "reservationAffinity": { "consumeReservationType": "NO_RESERVATION" }, "scheduling": { "instanceTerminationAction": "DELETE", "maxRunDuration": { "seconds": RUN_DURATION }, "onHostMaintenance": "TERMINATE", "provisioningModel": "FLEX_START" }

} }

Replace the following:

INSTANCE_TEMPLATE_NAME: the name of the instance template.
MACHINE_TYPE: the machine type to use for the VM. Specify either an A4 or A3 Ultra machine type. For more information, seeGPU machine types.
IMAGE_FAMILY: the image family of the OS image that you want to use. For a list of supported operating systems, see Supported operating systems.
IMAGE_PROJECT: the project ID of the OS image.
REGION: the region where you want to create the instance template. Specify a region in which the machine type that you want to use is available. For information about regions, seeGPU availability by regions and zones.
DISK_SIZE: the size of the boot disk in GB.
NETWORK_PROJECT_ID: the project ID of the network.
GVNIC_NAME_PREFIX: the name prefix that you specified when creating the standard VPC networks and subnets that use gVNIC NICs.
REGION: the region of the subnetwork.
RDMA_NAME_PREFIX: the name prefix that you specified when creating the VPC networks and subnets that use RDMA NICs.
RUN_DURATION: the duration, in seconds, you want the requested VMs to run. The value must be between 600, which is 600 seconds (10 minutes), and 604800, which is 604,800 seconds (seven days).

Reservation-bound

To create a regional instance template, make a POST request to the regionInstanceTemplates.insert method.

POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/instanceTemplates { "name":"INSTANCE_TEMPLATE_NAME", "properties":{ "machineType":"MACHINE_TYPE", "disks":[ { "boot":true, "initializeParams":{ "diskSizeGb":"DISK_SIZE", "diskType":"hyperdisk-balanced", "sourceImage":"projects/IMAGE_PROJECT/global/images/family/IMAGE_FAMILY" }, "mode":"READ_WRITE", "type":"PERSISTENT" } ], "serviceAccounts": [ { "email": "default", "scopes": [ "https://www.googleapis.com/auth/cloud-platform" ] } ], "networkInterfaces": [ { "accessConfigs": [ { "name": "external-nat", "type": "ONE_TO_ONE_NAT" } ], "network": "projects/NETWORK_PROJECT_ID/global/networks/GVNIC_NAME_PREFIX-net-0", "nicType": "GVNIC", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/GVNIC_NAME_PREFIX-sub-0" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/GVNIC_NAME_PREFIX-net-1", "nicType": "GVNIC", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/GVNIC_NAME_PREFIX-sub-1" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-0" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-1" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-2" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-3" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-4" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-5" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-6" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-7" } ], "reservationAffinity":{ "consumeReservationType":"SPECIFIC_RESERVATION", "key":"compute.googleapis.com/reservation-name", "values":[ "RESERVATION" ] }, "scheduling":{ "provisioningModel":"RESERVATION_BOUND", "instanceTerminationAction":"DELETE", "onHostMaintenance": "TERMINATE", "automaticRestart":true } } }

Replace the following:

INSTANCE_TEMPLATE_NAME: the name of the instance template.
MACHINE_TYPE: the machine type to use for the VM. Specify either an A4 or A3 Ultra machine type. For more information, seeGPU machine types.
IMAGE_FAMILY: the image family of the OS image that you want to use. For a list of supported operating systems, see Supported operating systems.
IMAGE_PROJECT: the project ID of the OS image.
REGION: the region where you want to create the instance template. Specify a region in which the machine type that you want to use is available. For information about regions, seeGPU availability by regions and zones.
DISK_SIZE: the size of the boot disk in GB.
NETWORK_PROJECT_ID: the project ID of the network.
GVNIC_NAME_PREFIX: the name prefix that you specified when creating the standard VPC networks and subnets that use gVNIC NICs.
REGION: the region of the subnetwork.
RDMA_NAME_PREFIX: the name prefix that you specified when creating the VPC networks and subnets that use RDMA NICs.
RESERVATION: either the reservation name or a specific block within a reservation. To get the reservation name or the available blocks, see View reserved capacity. Based on your requirement for instance placement, choose one of the following:
- To create instances across blocks or on a single block:
  projects/RESERVATION_OWNER_PROJECT_ID/reservations/RESERVATION_NAME
  Additionally, for a single block, create the MIG by applying a workload policy that specifies a block collocation (maxTopologyDistance=BLOCK) . Compute Engine then applies the policy to the reservation and creates instances on the same block.
- To create instances on a specific block:
  projects/RESERVATION_OWNER_PROJECT_ID/reservations/RESERVATION_NAME/reservationBlocks/RESERVATION_BLOCK_NAME

Spot

To create a regional instance template, make a POST request to the regionInstanceTemplates.insert method.

POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/instanceTemplates { "name":"INSTANCE_TEMPLATE_NAME", "properties":{ "machineType":"MACHINE_TYPE", "disks":[ { "boot":true, "initializeParams":{ "diskSizeGb":"DISK_SIZE", "diskType":"hyperdisk-balanced", "sourceImage":"projects/IMAGE_PROJECT/global/images/family/IMAGE_FAMILY" }, "mode":"READ_WRITE", "type":"PERSISTENT" } ], "serviceAccounts": [ { "email": "default", "scopes": [ "https://www.googleapis.com/auth/cloud-platform" ] } ], "networkInterfaces": [ { "accessConfigs": [ { "name": "external-nat", "type": "ONE_TO_ONE_NAT" } ], "network": "projects/NETWORK_PROJECT_ID/global/networks/GVNIC_NAME_PREFIX-net-0", "nicType": "GVNIC", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/GVNIC_NAME_PREFIX-sub-0" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/GVNIC_NAME_PREFIX-net-1", "nicType": "GVNIC", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/GVNIC_NAME_PREFIX-sub-1" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-0" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-1" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-2" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-3" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-4" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-5" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-6" }, { "network": "projects/NETWORK_PROJECT_ID/global/networks/RDMA_NAME_PREFIX-mrdma", "nicType": "MRDMA", "subnetwork": "projects/NETWORK_PROJECT_ID/region/REGION/subnetworks/RDMA_NAME_PREFIX-mrdma-sub-7" } ], "scheduling": { "provisioningModel": "SPOT", "instanceTerminationAction": "TERMINATION_ACTION", "onHostMaintenance": "TERMINATE", "automaticRestart": false } } }

Replace the following:

INSTANCE_TEMPLATE_NAME: the name of the instance template.
MACHINE_TYPE: the machine type to use for the VM. Specify either an A4 or A3 Ultra machine type. For more information, seeGPU machine types.
IMAGE_FAMILY: the image family of the OS image that you want to use. For a list of supported operating systems, see Supported operating systems.
IMAGE_PROJECT: the project ID of the OS image.
REGION: the region where you want to create the instance template. Specify a region in which the machine type that you want to use is available. For information about regions, seeGPU availability by regions and zones.
DISK_SIZE: the size of the boot disk in GB.
NETWORK_PROJECT_ID: the project ID of the network.
GVNIC_NAME_PREFIX: the name prefix that you specified when creating the standard VPC networks and subnets that use gVNIC NICs.
REGION: the region of the subnetwork.
RDMA_NAME_PREFIX: the name prefix that you specified when creating the VPC networks and subnets that use RDMA NICs.
TERMINATION_ACTION: the action to take when Compute Engine preempts the instance, either STOP (default) or DELETE.

After you create the instance template, you canview itto see its ID and review its instance properties.

Create a MIG

After you complete all the previous steps, create a MIG based on your scenario as follows:

Scenario	Method to create a MIG and VMs in it	Example
You have multiple or parallel jobs that can start with any number of VMs.	Create a MIG and use the target size to specify the number of VMs that you want in the group.See Create a MIG with a target size.	ML inference jobs
You have a job that requires distribution across an exact number of VMs.	Create a MIG without any VMs in it, and then create a resize request in the MIG. The resize request helps you to obtain VMs all at once. See Create a MIG and a resize request.	Distributed ML training and fine-tuning jobs

Create a MIG with a target size

If you can start your job without creating all of the VMs at once, then create a MIG with a target size. The target size determines the number of VMs in the MIG. The MIG starts creating VMs based on current resource availability. If any resources are temporarily unavailable, then the MIG continuously attempts to create VMs to meet the target size.

To create a MIG with a target size, select one of the following options:

gcloud

To create a MIG with a specified target size, use theinstance-groups managed create command.

The commands that you use to create a MIG use a workload policy to specify instance placement. If you don't want to use a workload policy, then remove the --workload-policy flag.

Create a zonal or regional MIG as follows:

To create a zonal MIG, use the following command:
gcloud compute instance-groups managed create MIG_NAME \
--template=INSTANCE_TEMPLATE_URL \
--size=TARGET_SIZE \
--workload-policy=WORKLOAD_POLICY_URL \
--zone=ZONE
To create a regional MIG, use the following command:
gcloud compute instance-groups managed create MIG_NAME \
--template=INSTANCE_TEMPLATE_URL \
--size=TARGET_SIZE \
--workload-policy=WORKLOAD_POLICY_URL \
--region=REGION Replace the following:
MIG_NAME: the name of the MIG.
INSTANCE_TEMPLATE_URL: the URL of the instance template that you want to use to create instances in the MIG. The URL can contain either theIDor name of the instance template. Specify one of the following values:
- For a regional instance template: projects/PROJECT_ID/regions/REGION/instanceTemplates/INSTANCE_TEMPLATE_ID
- For a global instance template: INSTANCE_TEMPLATE_ID
TARGET_SIZE: the number of instances that you want in the MIG.
WORKLOAD_POLICY_URL: Optional. The URL of the workload policy—for example,projects/example-project/regions/us-central1/resourcePolicies/example-workload-policy. If you don't want to use a workload policy, then remove the --workload-policy flag.
ZONE: the zone in which you want to create the MIG. If you use a workload policy, then specify a zone within the policy's region.
REGION: the region in which you want to create the MIG. If you use a workload policy, then specify the same region as that of the policy. For a regional MIG, instead of a region, you can specify the zones in that region by using the --zones flag.

REST

To create a MIG with a specified target size, make a POST request.

The requests that you use to create a MIG use a workload policy to specify instance placement. If you don't want to use a workload policy, then remove the resourcePolicies.workloadPolicyfield.

Create a zonal or regional MIG as follows:

To create a zonal MIG, make a POST request to theinstanceGroupManagers.insert method.
POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers
{
"versions": [
{
"instanceTemplate": "INSTANCE_TEMPLATE_URL"
}
],
"name": "MIG_NAME",
"targetSize": TARGET_SIZE,
"resourcePolicies": {
"workloadPolicy": "WORKLOAD_POLICY_URL"
}
}
To create a regional MIG, make a POST request to theregionInstanceGroupManagers.insert method.
POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/instanceGroupManagers
{
"name": "MIG_NAME",
"instanceTemplate": "INSTANCE_TEMPLATE_URL",
"targetSize": TARGET_SIZE,
"resourcePolicies": {
"workloadPolicy": "WORKLOAD_POLICY_URL"
}
} Replace the following:
PROJECT_ID: the project ID.
ZONE: the zone in which you want to create the MIG. If you use a workload policy, then specify a zone within the policy's region.
REGION: the region in which you want to create the MIG. If you use a workload policy, then specify the same region as that of the policy.
INSTANCE_TEMPLATE_URL: the URL of the instance template that you want to use to create instances in the MIG. The URL can contain either theIDor name of the instance template. Specify one of the following values:
- For a regional instance template: projects/PROJECT_ID/regions/REGION/instanceTemplates/INSTANCE_TEMPLATE_ID
- For a global instance template: INSTANCE_TEMPLATE_ID
MIG_NAME: the name of the MIG.
TARGET_SIZE: the number of instances that you want in the MIG.
WORKLOAD_POLICY_URL: Optional. The URL of the workload policy—for example,projects/example-project/regions/us-central1/resourcePolicies/example-workload-policy. If you don't want to use a workload policy, then remove the resourcePolicies.workloadPolicy field.

Create a MIG and a resize request

If you require multiple VMs all at once to start a job, then create a MIG, and create aresize request in the MIG as described in this section.

To create a resize request in a MIG, select one of the following options:

gcloud

The parameters that you need to specify depend on theconsumption option that you are using for this deployment. Select the tab that corresponds to your consumption option's provisioning model.

Flex-start

Create a zonal or regional MIG and a resize request as follows:

To create a zonal MIG and a resize request in it, do the following:
1. Create a zonal MIG using theinstance-groups managed create commandas follows.
  gcloud compute instance-groups managed create MIG_NAME \
  --template=INSTANCE_TEMPLATE_URL \
  --size=0 \
  --default-action-on-vm-failure=do-nothing \
  --workload-policy=WORKLOAD_POLICY_URL \
  --zone=ZONE
2. Create a resize request in the zonal MIG using theinstance-groups managed resize-requests create commandas follows. This command specifies the number of VMs that you want to create. If your workload requires specific VM names, thencreate a resize request by specifying a list of VM names.
  gcloud compute instance-groups managed resize-requests create MIG_NAME \
  --resize-request=RESIZE_REQUEST_NAME \
  --resize-by=COUNT \
  --zone=ZONE
To create a regional MIG and a resize request in it, do the following:
1. Create a regional MIG using theinstance-groups managed create commandas follows.
  gcloud compute instance-groups managed create MIG_NAME \
  --template=INSTANCE_TEMPLATE_URL \
  --size=0 \
  --default-action-on-vm-failure=do-nothing \
  --region=REGION \
  --target-distribution-shape=any-single-zone \
  --instance-redistribution-type=none \
  --workload-policy=WORKLOAD_POLICY_URL
2. Create a resize request in the regional MIG using thebeta instance-groups managed resize-requests create commandas follows. This command specifies the number of VMs that you want to create. If your workload requires specific VM names, thencreate a resize request by specifying a list of VM names.
  gcloud beta compute instance-groups managed resize-requests create MIG_NAME \
  --resize-request=RESIZE_REQUEST_NAME \
  --resize-by=COUNT \
  --region=REGION

Reservation-bound

The commands that you use to create a MIG use a workload policy to specify instance placement. If you don't want to use a workload policy, then remove the --workload-policy flag.

Create a zonal or regional MIG and a resize request as follows:

To create a zonal MIG and a resize request in it, do the following:
1. Create a zonal MIG using theinstance-groups managed create commandas follows.
  gcloud compute instance-groups managed create MIG_NAME \
  --template=INSTANCE_TEMPLATE_URL \
  --size=0 \
  --workload-policy=WORKLOAD_POLICY_URL \
  --zone=ZONE
2. Create a resize request in the zonal MIG using theinstance-groups managed resize-requests create commandas follows. This command specifies the number of VMs that you want to create. If your workload requires specific VM names, thencreate a resize request by specifying a list of VM names.
  gcloud compute instance-groups managed resize-requests create MIG_NAME \
  --resize-request=RESIZE_REQUEST_NAME \
  --resize-by=COUNT \
  --zone=ZONE
To create a regional MIG and a resize request in it, do the following:
1. Create a regional MIG using theinstance-groups managed create commandas follows.
  gcloud compute instance-groups managed create MIG_NAME \
  --template=INSTANCE_TEMPLATE_URL \
  --size=0 \
  --workload-policy=WORKLOAD_POLICY_URL \
  --zones=ZONE \
  --target-distribution-shape=any-single-zone \
  --instance-redistribution-type=none
2. Create a resize request in the regional MIG using thebeta instance-groups managed resize-requests create commandas follows. This command specifies the number of VMs that you want to create. If your workload requires specific VM names, thencreate a resize request by specifying a list of VM names.
  gcloud beta compute instance-groups managed resize-requests create MIG_NAME \
  --resize-request=RESIZE_REQUEST_NAME \
  --resize-by=COUNT \
  --region=REGION

Spot

The commands that you use to create a MIG use a workload policy to specify instance placement. If you don't want to use a workload policy, then remove the --workload-policy flag.

Create a zonal or regional MIG and a resize request as follows:

To create a zonal MIG and a resize request in it, do the following:
1. Create a zonal MIG using theinstance-groups managed create commandas follows.
  gcloud compute instance-groups managed create MIG_NAME \
  --template=INSTANCE_TEMPLATE_URL \
  --size=0 \
  --workload-policy=WORKLOAD_POLICY_URL \
  --zone=ZONE
2. Create a resize request in the zonal MIG using theinstance-groups managed resize-requests create commandas follows. This command specifies the number of VMs that you want to create. If your workload requires specific VM names, thencreate a resize request by specifying a list of VM names.
  gcloud compute instance-groups managed resize-requests create MIG_NAME \
  --resize-request=RESIZE_REQUEST_NAME \
  --resize-by=COUNT \
  --zone=ZONE
To create a regional MIG and a resize request in it, do the following:
1. Create a regional MIG using theinstance-groups managed create commandas follows.
  gcloud compute instance-groups managed create MIG_NAME \
  --template=INSTANCE_TEMPLATE_URL \
  --size=0 \
  --workload-policy=WORKLOAD_POLICY_URL \
  --zones=ZONE \
  --target-distribution-shape=any-single-zone \
  --instance-redistribution-type=none
2. Create a resize request in the regional MIG using thebeta instance-groups managed resize-requests create commandas follows. This command specifies the number of VMs that you want to create. If your workload requires specific VM names, thencreate a resize request by specifying a list of VM names.
  gcloud beta compute instance-groups managed resize-requests create MIG_NAME \
  --resize-request=RESIZE_REQUEST_NAME \
  --resize-by=COUNT \
  --region=REGION

Replace the following:

MIG_NAME: the name of the MIG.
INSTANCE_TEMPLATE_URL: the URL of the instance template that you want to use to create instances in the MIG. The URL can contain either theIDor name of the instance template. Specify one of the following values:
- For a regional instance template: projects/PROJECT_ID/regions/REGION/instanceTemplates/INSTANCE_TEMPLATE_ID
- For a global instance template: INSTANCE_TEMPLATE_ID
WORKLOAD_POLICY_URL: Optional. The URL of the workload policy—for example,projects/example-project/regions/us-central1/resourcePolicies/example-workload-policy. If you don't want to use a workload policy, then remove the --workload-policy flag.
ZONE: the zone in which you want to create the MIG. For a regional MIG also, you must specify a zone. This zone must be the zone that contains the profile for yourVPC network and must be azone where the machine type is available. For more information, see Limitations.
RESIZE_REQUEST_NAME: the name of the resize request, which must be unique within the specified MIG. Otherwise, creating the resize request fails.
COUNT: the number of instances to add to the MIG all at once.
REGION: the region in which you want to create the MIG.

Create a resize request by specifying a list of VM names

If your workload requires specific VM names, then specify a list of names to create the VMs. The number of names that you specify determines the number of VMs that the MIG creates all at once.

To create a resize request with specific VM names in a MIG, use thebeta gcloud compute instance-groups managed resize-requests create commandwith the --instances flag:

In a zonal MIG, run the following command:
gcloud beta compute instance-groups managed resize-requests create MIG_NAME \
--resize-request=RESIZE_REQUEST_NAME \
--instances=INSTANCE_NAMES \
--zone=ZONE
In a regional MIG, run the following command:
gcloud beta compute instance-groups managed resize-requests create MIG_NAME \
--resize-request=RESIZE_REQUEST_NAME \
--instances=INSTANCE_NAMES \
--region=REGION Replace INSTANCE_NAMES with a comma-separated list of names for the VMs. For example, instance-1,instance-2,instance-3.

REST

The parameters that you need to specify depend on theconsumption option that you are using for this deployment. Select the tab that corresponds to your consumption option's provisioning model.

Flex-start

Create a zonal or regional MIG and a resize request as follows:

To create a zonal MIG and a resize request in it, do the following:
1. Create a zonal MIG by making a POST request to theinstanceGroupManagers.insert method as follows.
  POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers
  {
  "versions": [
  {
  "instanceTemplate": "INSTANCE_TEMPLATE_URL"
  }
  ],
  "name": "MIG_NAME",
  "targetSize": 0,
  "instanceLifecyclePolicy": {
  "defaultActionOnFailure": "DO_NOTHING"
  },
  "resourcePolicies": {
  "workloadPolicy": "WORKLOAD_POLICY_URL"
  }
  }
2. Create a resize request in the zonal MIG by making a POST request to theinstanceGroupManagerResizeRequests.insert method as follows. This request specifies the number of VMs that you want to create. If your workload requires specific VM names, thencreate a resize request by specifying a list of VM names.
  POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers/MIG_NAME/resizeRequests
  {
  "name": "RESIZE_REQUEST_NAME",
  "resizeBy": COUNT
  }
To create a regional MIG and a resize request in it, do the following:
1. Create a regional MIG by making a POST request to theregionInstanceGroupManagers.insert method as follows.
  POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/instanceGroupManagers
  {
  "versions": [
  {
  "instanceTemplate": "INSTANCE_TEMPLATE_URL"
  }
  ],
  "name": "MIG_NAME",
  "targetSize": 0,
  "distributionPolicy": {
  "targetShape": "ANY_SINGLE_ZONE",
  "zones": [
  {
  "zone": "projects/PROJECT_ID/zones/ZONE"
  }
  ]
  },
  "updatePolicy": {
  "instanceRedistributionType": "NONE"
  },
  "instanceLifecyclePolicy": {
  "defaultActionOnFailure": "DO_NOTHING"
  },
  "resourcePolicies": {
  "workloadPolicy": "WORKLOAD_POLICY_URL"
  }
  }
2. Create a resize request in the regional MIG by making a POST request to thebeta.regionInstanceGroupManagerResizeRequests.insert method as follows. This request specifies the number of VMs that you want to create. If your workload requires specific VM names, thencreate a resize request by specifying a list of VM names.
  POST https://compute.googleapis.com/compute/beta/projects/PROJECT_ID/regions/REGION/instanceGroupManagers/MIG_NAME/resizeRequests
  {
  "name": "RESIZE_REQUEST_NAME",
  "resizeBy": COUNT
  }

Reservation-bound

The requests that you use to create a MIG use a workload policy to specify instance placement. If you don't want to use a workload policy, then remove the resourcePolicies.workloadPolicyfield.

Create a zonal or regional MIG and a resize request as follows:

To create a zonal MIG and a resize request in it, do the following:
1. Create a zonal MIG by making a POST request to theinstanceGroupManagers.insert method as follows.
  POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers
  {
  "versions": [
  {
  "instanceTemplate": "INSTANCE_TEMPLATE_URL"
  }
  ],
  "name": "MIG_NAME",
  "targetSize": 0,
  "resourcePolicies": {
  "workloadPolicy": "WORKLOAD_POLICY_URL"
  }
  }
2. Create a resize request in the zonal MIG by making a POST request to theinstanceGroupManagerResizeRequests.insert method as follows. This request specifies the number of VMs that you want to create. If your workload requires specific VM names, thencreate a resize request by specifying a list of VM names.
  POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers/MIG_NAME/resizeRequests
  {
  "name": "RESIZE_REQUEST_NAME",
  "resizeBy": COUNT
  }
To create a regional MIG and a resize request in it, do the following:
1. Create a regional MIG by making a POST request to theregionInstanceGroupManagers.insert method as follows.
  POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/instanceGroupManagers
  {
  "versions": [
  {
  "instanceTemplate": "INSTANCE_TEMPLATE_URL"
  }
  ],
  "name": "MIG_NAME",
  "targetSize": 0,
  "distributionPolicy": {
  "targetShape": "ANY_SINGLE_ZONE",
  "zones": [
  {
  "zone": "projects/PROJECT_ID/zones/ZONE"
  }
  ]
  },
  "updatePolicy": {
  "instanceRedistributionType": "NONE"
  },
  "resourcePolicies": {
  "workloadPolicy": "WORKLOAD_POLICY_URL"
  }
  }
2. Create a resize request in the regional MIG by making a POST request to thebeta.regionInstanceGroupManagerResizeRequests.insert method as follows. This request specifies the number of VMs that you want to create. If your workload requires specific VM names, thencreate a resize request by specifying a list of VM names.
  POST https://compute.googleapis.com/compute/beta/projects/PROJECT_ID/regions/REGION/instanceGroupManagers/MIG_NAME/resizeRequests
  {
  "name": "RESIZE_REQUEST_NAME",
  "resizeBy": COUNT
  }

Spot

The requests that you use to create a MIG use a workload policy to specify instance placement. If you don't want to use a workload policy, then remove the resourcePolicies.workloadPolicyfield.

Create a zonal or regional MIG and a resize request as follows:

To create a zonal MIG and a resize request in it, do the following:
1. Create a zonal MIG by making a POST request to theinstanceGroupManagers.insert method as follows.
  POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers
  {
  "versions": [
  {
  "instanceTemplate": "INSTANCE_TEMPLATE_URL"
  }
  ],
  "name": "MIG_NAME",
  "targetSize": 0,
  "resourcePolicies": {
  "workloadPolicy": "WORKLOAD_POLICY_URL"
  }
  }
2. Create a resize request in the zonal MIG by making a POST request to theinstanceGroupManagerResizeRequests.insert method as follows. This request specifies the number of VMs that you want to create. If your workload requires specific VM names, thencreate a resize request by specifying a list of VM names.
  POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers/MIG_NAME/resizeRequests
  {
  "name": "RESIZE_REQUEST_NAME",
  "resizeBy": COUNT
  }
To create a regional MIG and a resize request in it, do the following:
1. Create a regional MIG by making a POST request to theregionInstanceGroupManagers.insert method as follows.
  POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/instanceGroupManagers
  {
  "versions": [
  {
  "instanceTemplate": "INSTANCE_TEMPLATE_URL"
  }
  ],
  "name": "MIG_NAME",
  "targetSize": 0,
  "distributionPolicy": {
  "targetShape": "ANY_SINGLE_ZONE",
  "zones": [
  {
  "zone": "projects/PROJECT_ID/zones/ZONE"
  }
  ]
  },
  "updatePolicy": {
  "instanceRedistributionType": "NONE"
  },
  "resourcePolicies": {
  "workloadPolicy": "WORKLOAD_POLICY_URL"
  }
  }
2. Create a resize request in the regional MIG by making a POST request to thebeta.regionInstanceGroupManagerResizeRequests.insert method as follows. This request specifies the number of VMs that you want to create. If your workload requires specific VM names, thencreate a resize request by specifying a list of VM names.
  POST https://compute.googleapis.com/compute/beta/projects/PROJECT_ID/regions/REGION/instanceGroupManagers/MIG_NAME/resizeRequests
  {
  "name": "RESIZE_REQUEST_NAME",
  "resizeBy": COUNT
  }

Replace the following:

PROJECT_ID: the project ID.
ZONE: the zone in which you want to create the MIG. This zone must be the same as the zone that contains the profile for yourVPC network and must be azone where your chosen machine type is available. For more information, see Limitations.
REGION: the region in which you want to create the MIG. This region must be the same region that contains the profile for yourVPC network and must contain azone where your chosen machine type is available. For more information, see Limitations.
INSTANCE_TEMPLATE_URL: the URL of the instance template that you want to use to create instances in the MIG. The URL can contain either theIDor name of the instance template. Specify one of the following values:
- For a regional instance template: projects/PROJECT_ID/regions/REGION/instanceTemplates/INSTANCE_TEMPLATE_ID
- For a global instance template: INSTANCE_TEMPLATE_ID
MIG_NAME: the name of the MIG.
WORKLOAD_POLICY_URL: Optional. The URL of the workload policy—for example,projects/example-project/regions/us-central1/resourcePolicies/example-workload-policy. If you don't want to use a workload policy, then remove theresourcePolicies.workloadPolicy field.
RESIZE_REQUEST_NAME: the name of the resize request, which must be unique within the specified MIG. Otherwise, creating the resize request fails.
COUNT: the number of instances to add to the MIG all at once

Create a resize request by specifying a list of VM names

If your workload requires specific VM names, then specify a list of names to create the VMs. The number of names that you specify determines the number of VMs that the MIG creates all at once.

To create a resize request with specific VM names in a MIG, use one of the following methods:

In a zonal MIG, send a POST request by using thebeta.instanceGroupManagerResizeRequests.insert method:
POST https://compute.googleapis.com/compute/beta/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers/MIG_NAME/resizeRequests
In a regional MIG, send a POST request by using thebeta.regionInstanceGroupManagerResizeRequests.insert method:
POST https://compute.googleapis.com/compute/beta/projects/PROJECT_ID/regions/REGION/instanceGroupManagers/MIG_NAME/resizeRequests

For example, to specify two VM names, include the following in the request body:

{ "name": "RESIZE_REQUEST_NAME", "instances": [ { "name": "INSTANCE_NAME_1" }, { "name": "INSTANCE_NAME_2" } ] }

Replace INSTANCE_NAME_1, INSTANCE_NAME_2 with the names for the VMs.

What's next?

Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.

Last updated 2026-06-15 UTC.

Create an AI-optimized MIG with A4 or A3 Ultra machine type (original) (raw)