After you'vecreated a managed instance group (MIG),you might want to add or remove VMs from that group to meet your workload'sneeds. If you have reached the defaultlimitfor the maximum number of VMs that you can add to a MIG, and you need more VMs,you canincrease the size limit of the MIG.

To add or remove VMs in a MIG, you can do the following:

• Automatically add or remove VMs by setting the target size of the MIG:
  • Use autoscaling for stateless applications
  • Create a resize request in the MIG
  • Manually set the size of the MIG
• Manually add, remove or abandon specific instances:
  • Create instances with specific names
  • Delete specific instances
  • Abandon specific instances

Before you begin

• If you haven't already, set upauthentication. Authentication verifies your identity for access to Google Cloud services and APIs. To run code or samples from a local development environment, you can authenticate to Compute Engine by selecting one of the following options:

  Select the tab for how you plan to use the samples on this page:

  Console

  When you use the Google Cloud console to access Google Cloud services and APIs, you don't need to set up authentication.

  gcloud

  1. Install the Google Cloud CLI. After installation,initialize the Google Cloud CLI by running the following command:

     gcloudinit

     If you're using an external identity provider (IdP), you must first sign in to the gcloud CLI with your federated identity.

     Note: If you installed the gcloud CLI previously, make sure you have the latest version by runninggcloud components update.
  2. Set a default region and zone.

  Terraform

  To use the Terraform samples on this page in a local development environment, install and initialize the gcloud CLI, and then set up Application Default Credentials with your user credentials.

  Install the Google Cloud CLI.

  If you're using an external identity provider (IdP), you must first sign in to the gcloud CLI with your federated identity.

  Note: If you installed the gcloud CLI previously, make sure you have the latest version by runninggcloud components update.

  If you're using a local shell, then create local authentication credentials for your user account:

  gcloudauthapplication-defaultlogin

  You don't need to do this if you're using Cloud Shell.

  If an authentication error is returned, and you are using an external identity provider (IdP), confirm that you have signed in to the gcloud CLI with your federated identity.

  For more information, see Set up authentication for a local development environment.

  REST

  To use the REST API samples on this page in a local development environment, you use the credentials you provide to the gcloud CLI.

  Install the Google Cloud CLI.

  If you're using an external identity provider (IdP), you must first sign in to the gcloud CLI with your federated identity.

  Note: If you installed the gcloud CLI previously, make sure you have the latest version by runninggcloud components update.

  For more information, seeAuthenticate for using REST in the Google Cloud authentication documentation.

Set the group's target size

You can set a MIG's target size automatically by using an autoscaler or set thesize manually.

Automatically adding and removing VMs in a MIG

You can configure managed instance groups to automatically add or removeVMs based on their workloads. Your applications can gracefully handleincreases in traffic, and you can reduce your costs when the need forcompute resources is lower. To start scaling your managed instance groups,seeCreate a MIG with autoscaling.

For more information, seeAutoscaling groups of instances.

Create a resize request

You can use resize requests to add GPU VMs to a MIG all at once. This featurehelps you avoid charges for partial capacity while Compute Engineprovisions all resources. MIG resize requests use the flex-start orreservation-bound provisioning models. These models let you obtain vCPUs andGPUs at a discounted price.

For more information, seeAbout resize requests in a MIG.

Manually setting the size of a MIG

If a managed instance group is not already set toautomatically scale, you can resize the groupmanually to change the number of instances. If you increase thesize, the managed instance group uses the current instance template to add newinstances. If you decrease the size, the managed instance group deletesVMs from the group. The group deletes instances with acurrentActionofDELETING,CREATING, andRECREATING before it deletes instances that arerunning with no scheduled actions.

If the group is part of abackend service thathasenabled connection draining,it can take up to an additional 60 seconds after the connection drainingduration has elapsed before the VM instance is removed or deleted.

gcloud compute instance-groups managed resizeINSTANCE_GROUP_NAME \    --sizeNEW_SIZE \    --zoneZONE

POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers/INSTANCE_GROUP_NAME/resize?size=NEW_SIZE

After you make a request to resize a managed instance group, theVMs start or stop as soon as the system can provision or delete them. Thisprocess can take a significant amount of time depending on the number ofinstances in the group.Verify the status of instancesin your managed instance group.

Disabling creation retries mode during resize

This product is subject to the "Pre-GA Offerings Terms" in the General Service Terms section of theService Specific Terms. Pre-GA products are available "as is" and might have limited support. For more information, see thelaunch stage descriptions.

By default, if the initial creation of a VM instance fails,the managed instance group continuously retries to create the VM,until the VM is successfully created. However, if you don't wantautomatic creation retries, you can disable the creation retries mode when youresize the instance group.In this mode, the managed instance group attempts to create all instances onlyonce. If there is anerrorduring instance creation, the managed instance group gives up on this instance,removes it from thegroup'slist of managed instances,and decreases the target size of the managed instance group.

This mode applies only during the first creation attempt of a VM. If aVM is created successfully while this mode is enabled, the VMbehaves the same way as all other VMs created with a regular resizerequest. In particular, if a running VM dies unexpectedly at a later timeand needs to be recreated, this mode does not affect the recreation behavior inthat scenario.

Disabling the creation retries mode is especially useful in scenarios where youhave systems automatically creating groups of VMs where an exact numberof VMs is not required. You might prefer to stabilize quickly on the sizeof the managed instance group and be flexible in the number of VMs in thegroup, rather than to wait indefinitely until all the requested instances arecreated, which might be delayed temporarily or permanently because of quotaerrors or other unrelated issues.

Limitations

• You cannot disable creation retries for a regional MIG with anEVEN targetdistribution shape.
• For regional MIGs with aBALANCED orANY_SINGLE_ZONE target distributionshape, you must disableproactive instance redistributionbefore you can disable creation retries mode.
• You cannot disable creation retries forautoscaled MIGs.

To resize a managed instance group when creation retries mode is disabled, usethe Google Cloud CLI or REST.

gcloud beta compute instance-groups managed resizeINSTANCE_GROUP_NAME \    --sizeNEW_SIZE \--no-creation-retries \    [--regionREGION | --zoneZONE]

POST https://compute.googleapis.com/compute/beta/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers/INSTANCE_GROUP_NAME/resizeAdvanced{ "targetSize":NEW_SIZE,"noCreationRetries": true}

Add, remove or abandon specific instances

The following sections show how to add, remove, or abandon specific instances ina MIG.

Creating instances with specific names in MIGs

If your workload or orchestration system requires specific instance names, youcan add those instances to an existing MIG. This section shows how to addinstances based on availability. If resources are unavailable, then the MIG addsa portion of the instances and creates the remaining ones when resources becomeavailable. To add GPU instances with specific names all at once to a MIG, seeinsteadCreate a resize request in this document.

The names that you assign to these instances persist if the MIG recreates them.For more information about preserving the state of MIG instances, seestateful MIGs.

Limitations

If you create instances with specific names:

• You mustturn offautoscaling during the instance creation process.
• You must set the MIG's update policy to beopportunistic.
• If you are updating a regional MIG:
  • You must disableproactive instance redistribution.
  • You can't select which zone to create the VM in.
• You can't use duplicate VM instance names within a project.

You can use thegcloud CLIorRESTto add VMs with specific names to an existing MIG.

gcloud compute instance-groups managed create-instanceINSTANCE_GROUP_NAME \    --instanceINSTANCE_NAME \    [--regionREGION | --zoneZONE]

POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/instanceGroupManagers/INSTANCE_GROUP_NAME/createInstances{  "instances": [    {      "name": "INSTANCE_NAME_1"    },    {      "name": "INSTANCE_NAME_2"    }  ]}

Deleting specific instances from a group

You can delete individual VM instances in a managed instance group. Deletinginstances reduces the specifiedtargetSize of the instance group and removesthe VMs from anytarget pools of which theyare a member.

Deleting specific instances from a managed instance group doesn't change anyspecified autoscaler settings. If you delete instances from a managed instancegroup, the autoscaler might detect an increase in the workload on the otherinstances in the group and increase the group size back to its previous level.To prevent this,stop the autoscalerbefore attempting to delete the instances.

If the group is part of abackend service thathasenabled connection draining,it can take up to an additional 60 seconds after the connection drainingduration has elapsed before the VM instance is removed or deleted.

gcloud compute instance-groups managed delete-instancesINSTANCE_GROUP_NAME \    --instancesINSTANCE_NAME_1,INSTANCE_NAME_2 \     [--regionREGION | --zoneZONE]

gcloud compute instance-groups managed delete-instances example-mig \    --instances=example-instance-1,example-instance-2,example-instance-3,example-instance-4 \    --region=us-central1 \--skip-instances-on-validation-error

Updated [https://www.googleapis.com/compute/v1/projects/example-project/regions/us-central1/instanceGroupManagers/example-mig].PROJECT          ZONE           INSTANCE               STATUS   VALIDATION_ERRORexample-project  us-central1-a  example-instance-1     SKIPPED  Cannot delete instance https://www.googleapis.com/compute/v1/projects/example-project/zones/us-central1-a/instances/example-instance-1 that was already deleted.example-project  us-central1-a  example-instance-2     SKIPPED  Cannot delete instance https://www.googleapis.com/compute/v1/projects/example-project/zones/us-central1-a/instances/example-instance-2 that was already abandoned.example-project  us-central1-a  example-instance-3     SUCCESSexample-project  us-central1-a  example-instance-4     SKIPPED  Cannot find instance https://www.googleapis.com/compute/v1/projects/example-project/zones/us-central1-a/instances/example-instance-4.

gcloud compute instance-groups managed delete-instances example-mig \    --instances=example-instance-1,example-instance-2,example-instance-3,example-instance-4 \    --region=us-central1Updated [https://www.googleapis.com/compute/v1/projects/example-project/regions/us-central1/instanceGroupManagers/example-mig].PROJECT          ZONE             INSTANCE              STATUSexample-project  us-central1-a    example-instance-1    FAILexample-project  us-central1-a    example-instance-2    FAILexample-project  us-central1-a    example-instance-3    FAILexample-project  us-central1-a    example-instance-4    FAIL

POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/instanceGroupManagers/INSTANCE_GROUP_NAME/deleteInstances{  "instances": [    "zones/ZONE/instances/INSTANCE_NAME_1",    "zones/ZONE/instances/INSTANCE_NAME_2"   ]}

POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/instanceGroupManagers/INSTANCE_GROUP_NAME/deleteInstances{ "instances": [  "zones/ZONE/instances/INSTANCE_NAME_1",  "zones/ZONE/instances/INSTANCE_NAME_2" ],"skipInstancesOnValidationError": true}

After you make a request to delete instances in a managed instance group,the instances stop as soon as the system can delete them. This process can takea significant amount of time depending on the number of instances that youdelete from the group.Verify the status of the grouporverify the status of the managed instances.

Abandoning instances from a MIG

You can separate a VM instance from a managed instance group todebug issues with individual instances without affecting the group as a whole.Abandoning an instance from a group also removes the instance fromload balancers that were assigned to themanaged instance group.Target pools that weremanually assigned to specific individual instances are not removed.

You cannot add an abandoned VM instance back to the group.Abandoning instances reduces the specifiedtargetSize of the instance group,but doesn't change any specified autoscaler settings. Managed instance groupswith an autoscaler continue to add or remove instances automatically asnecessary. For a regional MIG with autoscaling set toONLY_SCALE_OUT mode,if you abandon a VM instance to reduce the MIG target size to theautoscaler's recommended size, thenthe autoscaler might create a new VM instance in the same zone of the abandonedVM. In such cases, to avoid creation of a new VM instance, try abandoning VMinstance from a different zone.

If the group is part of abackend service thathasenabled connection draining,it can take up to an additional 60 seconds after the connection drainingduration has elapsed before the VM instance is removed or deleted.

gcloud compute instance-groups managed abandon-instancesINSTANCE_GROUP_NAME \    --instancesINSTANCE_NAME_1,INSTANCE_NAME_2 \    [--regionREGION | --zoneZONE]

POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/instanceGroupManagers/INSTANCE_GROUP_NAME/abandonInstances{ "instances": [  "zones/ZONE/instances/INSTANCE_NAME_1",  "zones/ZONE/instances/INSTANCE_NAME_2" ]}

After you make a request to abandon instances from a managed instance group,the group removes the instances as soon as possible.Verify the status of the grouporverify the status of the managed instances.

Increase the group's size limit

By default, the size limit for a zonal MIG is 1,000 VMs, and the size limit fora regional MIG is 2,000 VMs. You can further increase these limits to 2,000 VMsfor a zonal MIG and 4,000 VMs for a regional MIG.

To increase the limits, set the group'slistManagedInstancesResults field toPAGINATED. If you still need more VMs in the group, create multiple MIGs orcontact support.

gcloud compute instance-groups managed updateINSTANCE_GROUP_NAME \--list-managed-instances-results=PAGINATED

gcloud compute instance-groups managed createINSTANCE_GROUP_NAME \    --templateINSTANCE_TEMPLATE \    --sizeSIZE \--list-managed-instances-results=PAGINATED

resource "google_compute_instance_group_manager" "default" {  name                           = "my-igm"  base_instance_name             = "test"  target_size                    = 7  zone                           = "us-central1-f"  list_managed_instances_results = "PAGINATED"  version {    instance_template = google_compute_instance_template.default.id    name              = "primary"  }}

PATCH https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/regions/REGION/regionInstanceGroupManagers/INSTANCE_GROUP_NAME{  "listManagedInstancesResults": "PAGINATED",}

If you want to switch back to pageless list results, you must also revertto the group's default size limit. To switch back, first ensure that thegroup does not exceed its default size limit, then set thelistManagedInstancesResults field toPAGELESS.

What's next

• Learn aboutworking with managed instances,for example, to delete, abandon, and recreate managed instances.
• View info aboutmanaged instance groups and managed instances.
• Troubleshoot managed instance groups.