Maintenance events and policies

To keep your Google Cloud Managed Lustre instances up-to-date, Google Cloud periodically performs maintenance on them. Maintenance events can include updates to the underlying software or infrastructure. These events may cause a temporary downtime, during which I/O operations to the instance may stop responding. Affected I/O operations resume once maintenance is complete. Downtime is typically between 2 to 25 minutes.

While maintenance is in progress, an instance's status is UPGRADING. Once maintenance has completed, the instance returns to ACTIVE status.

Maintenance policies allow you to control when these maintenance events occur. You can configure a maintenance policy for each instance to specify windows for maintenance and periods where maintenance should be excluded. If no maintenance policy is configured on an instance, maintenance can occur at any time.

You can also reschedule a maintenance event so that it starts immediately, during the next available window, or at a specific time up to 28 days later.

Extreme scale instances

Certain Managed Lustre configurations that provide extreme performance and scale may require planned downtime of up to 4 hours for periodic host maintenance. Google coordinates these maintenance events with you in advance.

About maintenance policies

You can configure a maintenance policy on each Managed Lustre instance. A maintenance policy includes the following:

Weekly maintenance window: A recurring one-hour window during the week when it is acceptable for Managed Lustre to start a maintenance event.
Maintenance exclusion window: A period during which Managed Lustre is not allowed to start a maintenance event. You can define one exclusion window per instance, with a maximum duration of two weeks. If no year is specified in the start and end dates, the exclusion window recurs annually.

An exclusion window takes precedence over a regular maintenance window. If a scheduled event falls within this period, Managed Lustre defers the event to the next available window or skips it entirely. The update is applied during the next maintenance cycle.

Maintenance notifications

You can subscribe to email notifications for upcoming maintenance events to help you plan for any potential disruption. Subscriptions to maintenance notifications are managed through your Google Cloud console Communication page:

Go to the Communication page

Once subscribed, the following notifications are sent:

Advance notification: Sent 7 to 14 days before a scheduled maintenance event, or after a maintenance event has been rescheduled by the customer.
Canceled: Sent if a scheduled maintenance event is canceled by Google or rescheduled by the customer.
In progress: Sent when a maintenance event starts.
Completed: Sent when a maintenance event is completed.

You can also view upcoming maintenance schedules on the instance details page in the Google Cloud console or you can use Cloud Logging to view maintenance logs.

Configure a maintenance window

A maintenance window requires a day of the week and a UTC start time for the one-hour window.

Google Cloud console

To configure a maintenance window when creating an instance, see Create an instance.

gcloud

Create an instance with a maintenance window by specifying the --maintenance-policy-weekly-windows flag:

gcloud lustre instances create INSTANCE_ID \
  --location=LOCATION \
  --network=NETWORK_NAME \
  --filesystem=FS_NAME \
  --capacity-gib=CAPACITY \
  --maintenance-policy-weekly-windows=dayOfWeek='DAY,startTime={START_TIME}'

Replace the following:

INSTANCE_ID: The ID of your Managed Lustre instance.
LOCATION: The location of your Managed Lustre instance.
NETWORK_NAME: The name of the network for your instance.
FS_NAME: The name of your file system.
CAPACITY: The capacity of your instance in GiB.
DAY: The day of the week for the maintenance window (e.g., TUESDAY).
START_TIME: The UTC start time of the maintenance window in dictionary format (e.g., {hours=1,minutes=0}).

REST API

To specify a maintenance window for an instance, include the maintenancePolicy object when creating an instance.

The following is an example of a maintenancePolicy object:

"maintenancePolicy": {
  "weeklyMaintenanceWindows": [
    {
      "dayOfWeek": "DAY",
      "startTime": {
        "hours": HOURS,
        "minutes": MINUTES
      }
    }
  ]
}

Replace the following:

DAY: The day of the week for the maintenance window (e.g., TUESDAY).
HOURS: The hour of the day for the maintenance window start time (e.g., 1).
MINUTES: The minute of the hour for the maintenance window start time (e.g., 0).

Configure an exclusion window

To prevent maintenance from occurring during certain periods, such as holidays or business events, you can define one maintenance exclusion window in your maintenance policy. Exclusion windows can be up to two weeks long. To create an annually-recurring exclusion window, don't specify a year in the start and end dates.

You can combine a weekly maintenance window and an exclusion window in the same policy.

Google Cloud console

To configure an exclusion window when creating an instance, see Create an instance.

gcloud

Create an instance with an exclusion window by specifying the --maintenance-policy-exclusion-window flag:

gcloud lustre instances create INSTANCE_ID \
  --location=LOCATION \
  --network=NETWORK_NAME \
  --filesystem=FS_NAME \
  --capacity-gib=CAPACITY \
  --maintenance-policy-exclusion-window=startDate=START_DATE,endDate=END_DATE,time={TIME}

Replace the following:

INSTANCE_ID: The ID of your Managed Lustre instance.
LOCATION: The location of your Managed Lustre instance.
NETWORK_NAME: The name of the network for your instance.
FS_NAME: The name of your file system.
CAPACITY: The capacity of your instance in GiB.
START_DATE: The start date of the maintenance exclusion window in YYYY-MM-DD format. To recur annually, use MM-DD format.
END_DATE: The end date of the maintenance exclusion window in YYYY-MM-DD format. To recur annually, use MM-DD format.
TIME: The UTC start and end time of the exclusion window in dictionary format (e.g., {hours=1,minutes=0}).

REST API

The following is an example of a maintenancePolicy object with an exclusion window:

"maintenancePolicy": {
  "maintenanceExclusionWindow": [
    {
      "startDate": {
        "year": START_YEAR,
        "month": START_MONTH,
        "day": START_DAY
      },
      "endDate": {
        "year": END_YEAR,
        "month": END_MONTH,
        "day": END_DAY
      },
      "time": {
        "hours": HOURS,
        "minutes": MINUTES
      }
    }
  ]
}

Replace the following:

START_YEAR: The year of the exclusion window start date (e.g., 2026). Omit this field and END_YEAR to create an annually-recurring window.
START_MONTH: The month of the exclusion window start date (e.g., 12).
START_DAY: The day of the exclusion window start date (e.g., 24).
END_YEAR: The year of the exclusion window end date (e.g., 2027). Omit this field and START_YEAR to create an annually-recurring window.
END_MONTH: The month of the exclusion window end date (e.g., 1).
END_DAY: The day of the exclusion window end date (e.g., 7).
HOURS: The hour of the day for the exclusion window time (e.g., 0).
MINUTES: The minute of the hour for the exclusion window time (e.g., 0).

Update a maintenance policy

To add or modify the maintenance policy for an existing instance:

Google Cloud console

To update the maintenance policy for an instance, see Manage instances.

gcloud

The following gcloud example updates the maintenance policy for an instance:

gcloud lustre instances update INSTANCE_ID \
--location=LOCATION \
--maintenance-policy-weekly-windows=dayOfWeek='DAY,startTime={START_TIME}'

Replace the following:

INSTANCE_ID: The ID of your Managed Lustre instance.
LOCATION: The location of your Managed Lustre instance.
DAY: The day of the week for the maintenance window (e.g., TUESDAY).
START_TIME: The UTC start time of the maintenance window in dictionary format (e.g., {hours=1,minutes=0}).

REST API

To specify a maintenance window for an instance, include the maintenancePolicy object when updating an instance. You must also specify the maintenancePolicy in the updateMask query parameter.

The following curl example updates the maintenance policy for an instance:

curl -X PATCH -H "Authorization: Bearer $(gcloud auth print-access-token)" -H "Content-Type: application/json" -d '{
  "maintenancePolicy": {
    "weeklyMaintenanceWindows": [
      {
        "dayOfWeek": "DAY",
        "startTime": {
          "hours": HOURS,
          "minutes": MINUTES
        }
      }
    ]
  }
}' "https://lustre.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/instances/INSTANCE_ID?updateMask=maintenancePolicy"

Replace the following:

PROJECT_ID: The ID of your Google Cloud project.
LOCATION The location of the Managed Lustre instance.
INSTANCE_ID The ID of the Managed Lustre instance.
DAY: The day of the week for the maintenance window (e.g., TUESDAY).
HOURS The hour of the day for the maintenance window start time (e.g., 3).
MINUTES The minute of the hour for the maintenance window start time (e.g., 45).

Remove a maintenance policy

To remove a maintenance policy from an instance, clear all weekly maintenance windows and exclusion windows from the policy.

Google Cloud console

To remove a maintenance policy from an instance, edit the instance and clear the maintenance policy fields. See Manage instances.

gcloud

Use the gcloud lustre instances update command with the --clear-maintenance-policy-weekly-windows and --clear-maintenance-policy-exclusion-window flags:

gcloud lustre instances update INSTANCE_ID \
--location=LOCATION \
--clear-maintenance-policy-weekly-windows \
--clear-maintenance-policy-exclusion-window

Replace the following:

INSTANCE_ID: The ID of your Managed Lustre instance.
LOCATION: The location of your Managed Lustre instance.

REST API

To remove a maintenance policy using the REST API, use the instances.patch method with an empty maintenancePolicy object and the updateMask query parameter set to maintenancePolicy:

curl -X PATCH -H "Authorization: Bearer $(gcloud auth print-access-token)" -H "Content-Type: application/json" -d '{
  "maintenancePolicy": {}
}' "https://lustre.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/instances/INSTANCE_ID?updateMask=maintenancePolicy"

Replace the following:

PROJECT_ID: The ID of your Google Cloud project.
LOCATION: The location of the Managed Lustre instance.
INSTANCE_ID: The ID of the Managed Lustre instance.

View maintenance logs

Managed Lustre uses Cloud Logging to log maintenance events, including the following activity types:

LustreAdvancedNotification: Maintenance has been scheduled for the instance.
LustreOngoingNotification: Maintenance has started on the instance. Once complete, another LustreOngoingNotification event is logged, this time with is_completed=true.
LustreCancelledNotification: Maintenance has been canceled.

You can view maintenance logs using Logs Explorer or by using Google Cloud CLI.

Logs Explorer

To view Managed Lustre maintenance logs using Logs Explorer, do the following:

In the Google Cloud console, go to the Cloud Logging > Logs Explorer page:

Go to Logs Explorer

Enter the following filter into the query editor and click Run query:

resource.type="lustre.googleapis.com/Instance"
resource.labels.instance_id="INSTANCE_ID"
resource.labels.location="LOCATION"
logName="projects/PROJECT_ID/logs/lustre.googleapis.com%2Fmaintenance"

gcloud CLI

To view maintenance logs with Google Cloud CLI, run the gcloud logging read command:

gcloud logging read 'resource.type="lustre.googleapis.com/Instance" resource.labels.instance_id="INSTANCE_ID" resource.labels.location="LOCATION" logName="projects/PROJECT_ID/logs/lustre.googleapis.com%2Fmaintenance"' --project=PROJECT_ID

Replace the following:

INSTANCE_ID: The ID of the Managed Lustre instance.
LOCATION: The zone where the instance is located.
PROJECT_ID: The project ID of the project where the instance is located.

Reschedule maintenance

If a scheduled maintenance event conflicts with your needs, you can reschedule it. When rescheduling, you have the following options:

Immediately: Reschedule the maintenance event to start as soon as possible. This option ignores any maintenance windows and exclusion windows that are configured on the instance.
Next available window: Reschedule the maintenance event to the next available maintenance window. This typically shifts the update by one week.
By date: Reschedule the maintenance event to a specific date and time. You can delay maintenance by a maximum of 28 days from the originally scheduled time. This option ignores any maintenance windows and exclusion windows that are configured on the instance.

To reschedule maintenance:

gcloud

To start the maintenance event as soon as possible:

gcloud lustre instances reschedule-maintenance INSTANCE_ID \
--location=LOCATION \
--reschedule-type=immediately

To reschedule maintenance to the next available window:

gcloud lustre instances reschedule-maintenance INSTANCE_ID \
--location=LOCATION \
--reschedule-type=next-available-window

To reschedule maintenance to a specific time:

gcloud lustre instances reschedule-maintenance INSTANCE_ID \
--location=LOCATION \
--reschedule-type=by-time \
--reschedule-schedule-time=TIMESTAMP

Replace the following:

INSTANCE_ID: The ID of your Managed Lustre instance.
LOCATION: The location of your Managed Lustre instance.
TIMESTAMP: The start time in RFC 3339 format (for example, 2026-04-30T10:00:00Z). This must be no later than 28 days from the originally scheduled time.

REST API

Use the instances.rescheduleMaintenance method.

The following curl example starts the maintenance event as soon as possible:

curl -X POST -H "Authorization: Bearer $(gcloud auth print-access-token)" -H "Content-Type: application/json" -d '{
  "reschedule": {
    "rescheduleType": "IMMEDIATELY"
  }
}' "https://lustre.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/instances/INSTANCE_ID:rescheduleMaintenance"

The following curl example reschedules maintenance to the next available window:

curl -X POST -H "Authorization: Bearer $(gcloud auth print-access-token)" -H "Content-Type: application/json" -d '{
  "reschedule": {
    "rescheduleType": "NEXT_AVAILABLE_WINDOW"
  }
}' "https://lustre.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/instances/INSTANCE_ID:rescheduleMaintenance"

The following curl example reschedules maintenance to a specific time:

curl -X POST -H "Authorization: Bearer $(gcloud auth print-access-token)" -H "Content-Type: application/json" -d '{
  "reschedule": {
    "rescheduleType": "BY_TIME",
    "scheduleTime": "TIMESTAMP"
  }
}' "https://lustre.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/instances/INSTANCE_ID:rescheduleMaintenance"

Replace the following:

PROJECT_ID: The ID of your Google Cloud project.
INSTANCE_ID: The ID of your Managed Lustre instance.
LOCATION: The location of your Managed Lustre instance.
TIMESTAMP: The start time in RFC 3339 format (for example, 2026-04-30T10:00:00Z). This must be no later than 28 days from the originally scheduled time.

Maintenance events and policies Stay organized with collections Save and categorize content based on your preferences.

Extreme scale instances

About maintenance policies

Maintenance notifications

Configure a maintenance window

Google Cloud console

gcloud

REST API

Configure an exclusion window

Google Cloud console

gcloud

REST API

Update a maintenance policy

Google Cloud console

gcloud

REST API

Remove a maintenance policy

Google Cloud console

gcloud

REST API

View maintenance logs

Logs Explorer

gcloud CLI

Reschedule maintenance

gcloud

REST API

Maintenance events and policies