Applying, viewing, and removing stateful configuration in MIGs

After you create or update a stateful configuration for a stateful managed instance group (MIG), you can:

  • Apply the stateful configuration for it to take effect.
  • View the stateful configuration as well as the effective preserved state of your managed instances.
  • Remove the stateful configuration.

Before you begin

Applying stateful configuration to managed instances

Your stateful configuration is effective after you or the MIG applies it.

  • Configuration changes in a stateful policy are automatically applied to all managed instances.
  • Configuration changes in a per-instance configuration can be applied manually or automatically.

For more conceptual information, read How stateful configuration is applied to managed instances.

Applying configuration from a stateful policy

All configuration changes in a stateful policy are automatically applied to all managed instances. Updates to a stateful policy do not disrupt running virtual machine (VM) instances.

Verifying whether a stateful policy has been applied

Verifying all VMs

Verify whether changes to a stateful policy have been applied to all VMs by checking if the MIG has become stable after the changes.

Verifying specific VMs

Verify if the changes to a stateful policy have been applied to a specific VM in a MIG by viewing the preserved state of the managed instance and checking if all disks from the stateful policy are present in either preservedStateFromPolicy or preservedStateFromConfig for the managed instance.

Applying stateful configuration from per-instance configurations

You can apply new or updated per-instance configurations either manually or automatically. Use one of the following approaches:

  • Selectively apply updated configurations to specific instances: Use this approach to control the disruption, the timing, and the sequence of the updates.
  • Automatically apply updated configurations with rolling updates: Use this approach to apply configuration changes to managed instances in an automated, rolling fashion.

Both of these methods apply outstanding updates from both per-instance configurations and unapplied versions or instance template, if relevant.

Selectively apply updated configuration to specific instances

To manually apply per-instance configurations to specific instances, use the following steps:

  1. Configure an opportunistic update policy to prevent a race with automatic proactive updates.
  2. Create or update per-instance configurations.
  3. Apply updated configurations to specific VMs by selectively updating those VMs.

Alternatively, you can use the following approaches:

  • When creating or updating a per-instance configuration with the gcloud tool, use the --update-instance flag to immediately apply the configuration to the associated managed instance.
  • When you manually create new instances, specify their names and per-instance configurations at creation time. The MIG applies configurations immediately on VM creation.

Automatically apply updated configuration with rolling updates

Configure proactive rolling updates for your MIG to automatically apply configuration updates to corresponding managed instances in a rolling fashion.

Stateful MIGs require the following configuration for a proactive update policy:

This configuration ensures that the MIG recreates existing stateful instances and does not substitute them with different ones.

You can also ensure that changes in per-instance configurations are applied without stopping instances, if possible, by setting the minimal action to REFRESH in the group's update policy.

Verifying whether per-instance configurations have been applied

Verifying all per-instance configurations

To verify whether all per-instance configurations in a MIG have been applied, view the MIG's stateful status and check status.stateful.perInstanceConfigs.allEffective:

  • true: All per-instance configurations in the group have been applied and are effective, or the group has no per-instance configurations.
  • false: The group has at least one per-instance configuration that is not yet effective: either you have not yet applied it or it is in the process of being applied.

Verifying a specific per-instance configuration

To verify if the changes to a specific per-instance configuration have been applied to its corresopnding VM, view all per-instance configurations and check the specific per-instance configuration's status field:

  • UNAPPLIED: The per-instance configuration was created or updated and you need to apply it to the VM with a manual instance update.
  • APPLYING: The MIG is currently applying the new or updated per-instance configuration to the VM.
  • EFFECTIVE: The per-instance configuration has been applied successfully to the VM and is effective.
  • UNAPPLIED_DELETION: The per-instance configuration is set to be deleted. You must apply this update to the VM with a manual instance update for it to take effect.
  • DELETING: The per-instance configuration is being deleted and the change is being applied to the VM.

Viewing stateful configuration and preserved state

Get information about your stateful MIGs for the following tasks:

  • Verifying whether a MIG has stateful configuration and whether this configuration is applied and effective.
  • Viewing stateful configuration that is common to all instances in the MIG, stored in the stateful policy.
  • Viewing stateful configuration that is specific to individual instances, stored in per-instance configurations, and checking whether this configuration is applied and effective.
  • Viewing the effective preserved state for each instance in a MIG based on its stateful policy and applied per-instance configurations.

Viewing the status of a MIG's stateful configuration

A MIG is considered stateful if it has any stateful configuration, that is, either a stateful policy or at least one non-empty per-instance configuration.

If you create a stateful policy, the MIG automatically applies it to make it effective. If you create per-instance configurations, you might decide to apply them later.

To verify whether a MIG has stateful configuration and that all per-instance configurations are applied, view the MIG's stateful status by using the gcloud tool or the Compute Engine API.

gcloud

To verify whether a MIG has stateful configuration, view its stateful status by running the gcloud compute instance-groups managed describe command. For example:

gcloud compute instance-groups managed describe NAME

baseInstanceName node
...
name example-group
...
status:
  isStable: true
  stateful:
    hasStatefulConfig: true
    perInstanceConfigs:
      allEffective: true
  versionTarget:
    isReached: true
targetSize: 3
...

In this output:

  • hasStatefulConfig:
    • true: The group has stateful configuration, that is, the group has a non-empty stateful policy or at least one non-empty per-instance configuration.
    • false: No stateful configuration exists.
  • perInstanceConfigs.allEffective:
    • true: All per-instance configurations in the group have been applied and are effective, or the group has no per-instance configurations.
    • false: The group has at least one per-instance configuration that is not yet effective: either you have not yet applied it or it is in the process of being applied.

API

To verify whether a MIG has stateful configuration, use the instanceGroupManagers.get or regionInstanceGroupManagers.get method and check the status.stateful field. For example:

GET https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers/NAME

The method returns the instanceGroupManagers object:

{
  "name": "example-group",
  "baseInstanceName": "node",
  ...
  "status": {
    "isStable": true,
    "stateful": {
      "hasStatefulConfig": true,
      "perInstanceConfigs": {
        "allEffective": true
      }
    },
    "versionTarget": {
      "isReached": true
    }
  },
  "targetSize": 3,
  ...
}

In this output:

  • hasStatefulConfig:
    • true: The group has stateful configuration, that is, the group has a non-empty stateful policy or at least one non-empty per-instance configuration.
    • false: No stateful configuration exists.
  • perInstanceConfigs.allEffective:
    • true: All per-instance configurations in the group have been applied and are effective, or the group has no per-instance configurations.
    • false: The group has at least one per-instance configuration that is not yet effective: either you have not yet applied it or it is in the process of being applied.

Viewing a MIG's stateful policy

View a MIG's stateful policy by checking the group's details with the gcloud tool or the Compute Engine API.

gcloud

To view a MIG's stateful policy, run the gcloud compute instance-groups managed describe command. For example:

gcloud compute instance-groups managed describe NAME

baseInstanceName node
...
name example-group
...
statefulPolicy:
  preservedState:
    disks:
      data-disk:
        autoDelete: NEVER
      logs-disk:
        autoDelete: ON_PERMANENT_INSTANCE_DELETION
status:
  isStable: true
  stateful:
    hasStatefulConfig: true
    perInstanceConfigs:
      allEffective: true
  versionTarget:
    isReached: true
targetSize: 3
...

API

To view a MIG's stateful policy use the instanceGroupManagers.get or regionInstanceGroupManagers.get method and check the statefulPolicy field. For example:

GET https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers/NAME

The method returns the instanceGroupManagers object:

{
  "name": "example-group",
  "baseInstanceName": "node",
  ...
  "status": {
      "isStable": true,
      "versionTarget": {
        "isReached": true
      },
      "stateful": {
        "hasStatefulConfig": true,
        "perInstanceConfigs": {
          "allEffective": true
        }
      }
  ...
  "statefulPolicy": {
    "preservedState": {
      "disks": {
        "data-disk": { "autoDelete": "NEVER" },
        "logs-disk": { "autoDelete": "ON_PERMANENT_INSTANCE_DELETION" }
      }
    }
  },
  "targetSize": 3,
  ...
}

Viewing a MIG's per-instance configurations

View per-instance configurations by listing all per-instance configurations in a MIG. Use the gcloud tool or the Compute Engine API.

If you decide to apply per-instance configurations manually, some per-instance configurations might not yet be applied to the associated instances and, therefore, the instances' preserved states might not yet reflect your per-instance configurations. For more information, see Applying per-instance configurations and Verifying whether per-instance configurations have been applied

gcloud

List all per-instance configurations in a MIG by running the gcloud compute instance-groups managed instance-configs list command:

gcloud compute instance-groups managed instance-configs list NAME

Replace the following:

  • NAME: the name of a MIG to list per-instance configurations from

Filter the list using the standard --filter flag.

For example, to list all per-instance configurations in example-group:

gcloud compute instance-groups managed instance-configs list example-group
---
fingerprint: JxPvsKOywuY=
name: node-1
preservedState:
  disks:
    data-disk:
      autoDelete: NEVER
      mode: rw
      source: https://www.googleapis.com/compute/v1/projects/example-project/zones/us-east1-c/disks/data-disk-1
  metadata:
    role: primary
status: EFFECTIVE
---
fingerprint: IbGmJBqqEkk=
name: node-2
preservedState:
...

API

List all per-instance configurations in a MIG by calling the instanceGroupManagers.listPerInstanceConfigs or regionInstanceGroupManagers.listPerInstanceConfigs method:

POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers/NAME/listPerInstanceConfigs

Replace the following:

  • PROJECT_ID: the project ID for the request
  • ZONE: the zone where the MIG is located (applies to a zonal MIG)
    • For a regional MIG, replace zones/ZONE with regions/REGION and specify the region of the MIG
  • NAME: the name of a MIG to list per-instance configurations from

For example, to list all per-instance configurations in example-group:

POST https://compute.googleapis.com/compute/v1/projects/example-project/zones/us-east1-c/instanceGroupManagers/example-group/listPerInstanceConfigs

The method returns the list of per-instance configurations in the group:

{
  "items": [
    {
      "fingerprint": "JxPvsKOywuY=",
      "preservedState" : {
        "disks": {
          "data-disk" : {
            "source": "https://www.googleapis.com/compute/v1/projects/example-project/zones/us-east1-c/disks/data-disk-1",
            "mode": "READ_WRITE",
            "autoDelete": "NEVER"
          }
        },
        "metadata": {
          "role": "primary"
        }
      },
      "name": "node-1",
      "status": "EFFECTIVE"
    },
    {
      "fingerprint": "IbGmJBqqEkk="
      "preservedState" : {
        ...
      }
      "name": "node-2",
      ...
    },
    ...
  ]
}

Viewing the preserved states of managed instances

When stateful configuration is applied, the MIG generates and sets the effective preserved state of each managed instance in two fields:

  • preservedStateFromPolicy: Contains the preserved state that is generated based on a stateful policy, excluding any stateful configuration that is overridden by per-instance configurations.
  • preservedStateFromConfig: Contains the preserved state that is generated based on a per-instance configuration that has already been applied to the managed instance.

To see the effective preserved state of each managed instance in a MIG, list the managed instances using the gcloud tool or the Compute Engine API.

gcloud

Check which managed instances have preserved state by using the gcloud compute instance-groups managed list-instances command and viewing values in the PRESERVED_STATE column. For example:

gcloud compute instance-groups managed list-instances NAME

NAME    ZONE        STATUS   HEALTH_STATE  ACTION  PRESERVED_STATE  INSTANCE_TEMPLATE  ...
node-1  us-east1-c  RUNNING                NONE    POLICY,CONFIG    example-template
node-2  us-east1-c  RUNNING                NONE    POLICY,CONFIG    example-template

The PRESERVED_STATE column can contain the following values:

  • POLICY: The managed instance has preserved state based on a stateful policy.
  • CONFIG: The managed instance has preserved state based on a per-instance configuration.
  • No value: The manage instance has no preserved state and is stateless.

View the preserved state of a specific managed instance by running the gcloud compute instance-groups managed describe-instance command:

gcloud compute instance-groups managed describe-instance INSTANCE_GROUP_NAME \
  --instance INSTANCE_NAME

Replace the following:

  • INSTANCE_GROUP_NAME: the name of a MIG
  • INSTANCE_NAME: the name of a managed instance in the group

For example, to view the preserved state of the node-1 managed instance in example-group, run:

gcloud compute instance-groups managed describe-instance example-group \
  --instance node-1
instance: .../example-project/zones/us-east1-c/instances/node-1
instanceStatus: RUNNING
currentAction: NONE
id: 123456789012345678
version:
  instanceTemplate: .../example-project/global/instanceTemplates/example-template
preservedStateFromPolicy:
  disks:
    data-disk:
      autoDelete: NEVER
      mode: rw
      source: .../example-project/zones/us-east1-c/disks/data-disk-1
preservedStateFromConfig:
  metadata:
    role: primary
    my-key: my-value

API

List managed instances to view each instance's preserved state by calling the instanceGroupManagers.listManagedInstances or regionInstanceGroupManagers.listManagedInstances method.

For example, to list all managed instances:

POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers/NAME/listManagedInstances

The method returns the list of managed instances in the group, which contains information about their preserved states:

{
  "managedInstances": [
    {
      "instance": ".../example-project/zones/us-east1-c/instances/node-1",
      "instanceStatus": "RUNNING",
      "currentAction": "NONE",
      "id": "123456789012345678",
      "version": {
         "instanceTemplate":".../example-project/global/instanceTemplates/example-template"
      },
      "preservedStateFromPolicy": {
        "disks": {
          "data-disk" : {
            "source": "https://www.googleapis.com/compute/v1/projects/example-project/zones/us-east1-c/disks/data-disk-1",
            "mode": "rw",
            "autoDelete": "NEVER"
          }
        }
      },
      "preservedStateFromConfig": {
        "metadata": {
          "role": "primary",
          "my-key": "my-value"
        }
      }
    },
    {
      "instance": ".../example-project/zones/us-east1-c/instances/node-2",
      "instanceStatus": "RUNNING",
      ...
      "preservedStateFromPolicy": {
        ...
      },
      "preservedStateFromConfig": {
        ...
      }
    },
    ...
  ]
}

Removing stateful configuration

Removing stateful configuration is useful in the following scenarios:

  • You have migrated standalone VMs to a stateful MIG and now want to move the stateful configuration from per-instance configurations to a common stateful policy.
  • You have rearchitected your workload and no longer need to keep any state on the VM instances.
  • You created a stateful configuration for testing purposes and now want to clean it up.

This section describes how to fully remove a stateful policy on a per-instance configuration. If you only need to remove a subset of a stateful policy or per-instance configuration, see the following sections:

Removing a stateful policy

When you remove a stateful policy, you affect all VMs in a MIG. The MIG treats all stateful resources that were configured in the stateful policy as stateless, unless those resources are also configured in per-instance configurations for individual instances. On subsequent instance recreation, autohealing, or update operations, the now-stateless resources can lose their state:

  • Disks can be deleted and recreated from the source, which is defined in the instance template.

For conceptual information, see how removing a resource from stateful policy affects preserved state.

gcloud

To delete all configuration from a stateful policy, run the gcloud compute instance-groups managed update command with the --remove-stateful-disks flag, and list the device names of all configured stateful disks.

For example, if your stateful policy contains configuration for two stateful disks with device names data-disk and logs-disk, run the following command to clear the policy:

gcloud compute instance-groups managed update NAME \
    --remove-stateful-disks data-disk,logs-disk

The MIG removes the data-disk and logs-disk configuration from stateful policy and removes the disks from the preserved states of all managed instances in the group automatically and asynchronously, unless the disks are also configured in per-instance configurations.

API

To delete all configuration in stateful policy, set the statefulPolicy field to null by using the instanceGroupManagers.patch. or regionInstanceGroupManagers.patch. method.

For example, the following call removes stateful policy configuration:

PATCH https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers/NAME

{
  "statefulPolicy": null
}

The MIG clears stateful policy configuration and removes the stateful resources from the preserved states of all managed instances in the group automatically and asynchronously, unless the resources are also configured in per-instance configurations.

Removing stateful configuration for a specific VM

Removing stateful configuration for a specific VM is done in two steps:

  1. Delete the associated per-instance configuration.
  2. Apply the change to the managed instance. You can choose whether the change should be applied manually or automatically.

After the change is applied, all stateful items that were previously configured in the per-instance configuration are now treated as stateless, unless they are also configured in the group's stateful policy. When the VM is subsequently recreated or updated, the state of the items is lost:

  • Disks can be detached or recreated from their source in the instance template.
  • Metadata is removed or reset to the values defined in the instance template.

For more information, see how removing items from per-instance configurations affects preserved state.

gcloud

To fully delete the per-instance configuration, run the gcloud compute instance-groups managed instance-configs delete command.

Use the optional --update-instance flag to apply the changes immediately to the instance (the default). If you set the --no-update-instance flag, the changes are applied when you next recreate or update the instance.

The --instance-update-minimal-action flag can only be used together with the --update-instance flag.

gcloud compute instance-groups managed instance-configs delete INSTANCE_GROUP_NAME \
  --instances INSTANCE_NAME[,INSTANCE_NAME,...] \
  [--no-update-instance | --update-instance] \
  [--instance-update-minimal-action MINIMAL_ACTION

Replace the following:

  • INSTANCE_GROUP_NAME: The name of the MIG.
  • INSTANCE_NAME: Names of the instances for which to delete per-instance configurations
  • MINIMAL_ACTION: The minimal action to perform when applying the per-instance configuration update to the VM. This value must be one of the following:
    • none: No action.
    • refresh: Apply updates that are possible to apply without stopping the VM.
    • restart: Stop the VM and then start it again.
    • replace: Recreate the VM.

For example, to delete the per-instance configuration for node-1 in example-group, run the following command:

gcloud compute instance-groups managed instance-configs delete example-group \
  --instances node-1 \
  --update-instance

The MIG deletes the per-instance configuration for node-1 and, because the VM was updated and configuration changes applied, the MIG no longer treats those previously stateful items as stateful.

API

To fully delete the per-instance configuration for one or multiple instances, use the instanceGroupManagers.deletePerInstanceConfigs or regionInstanceGroupManagers.deletePerInstanceConfigs method.

POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/zones/ZONE/instanceGroupManagers/INSTANCE_GROUP_NAME/deletePerInstanceConfigs

{
  "names": ["INSTANCE_NAME",...]
}

Replace the following:

  • PROJECT_ID: the project ID for the request
  • ZONE: the zone where the MIG is located (applies to a zonal MIG)
    • For a regional MIG, replace zones/ZONE with regions/REGION and specify the region of the MIG
  • INSTANCE_GROUP_NAME: the name of the MIG
  • INSTANCE_NAME: names of the instances for which to delete per-instance configurations

The deletePerInstanceConfig method deletes the specified per-instance configurations but does not apply the changes to the associated VMs. The changes are applied to a VM when you recreate or update the instance. Update selected VMs manually to apply the changes.

For example, to delete per-instance configuration for node-1 in example-group, call the following method:

POST https://compute.googleapis.com/compute/v1/projects/example-project/zones/us-east1-c/instanceGroupManagers/example-group/deletePerInstanceConfigs

{
  "names": ["node-1"]
}

The method deletes per-instance configuration for the node-1 instance from example-group. The change will be applied to the managed instance when you recreate or update the instance.

To apply the configuration change, call the instanceGroupManagers.applyUpdatesToInstances method:

POST https://compute.googleapis.com/compute/v1/projects/example-project/zones/us-east1-c/instanceGroupManagers/example-group/applyUpdatesToInstances

{
  "instances": ["/zones/us-east1-c/instances/node-1"]
}

The method removes stateful items from the managed instance's preserved state.

What's next