Idle shutdown

Vertex AI Workbench instancesshut down after a specified period of inactivity by default.This page describes the idle shutdown feature and how tochange the default idle shutdown settings during instance creation.

Overview

To help manage costs, Vertex AI Workbench instancesshut down after being idle for a specific time period by default.You can change the amount of time or turn this feature off.

Requirements for running idle shutdown

For idle shutdown to run, your Vertex AI Workbench instancemust have guest attributes enabled. Guest attributes are enabled by defaultbut if you've turned off guest attributes, you can enable guest attributesby setting theenable-guest-attributes metadata key totrue. SeeUpdate an instance's metadata.

Billing

While your instance is shut down, there are no CPU or GPUusage charges except for scheduled executions that run duringthe shutdown. For more information about scheduled executions, seeScheduled executions run while instance is shut downon this page.

Disk storage charges still apply whileyour instance is shut down. For more information,seePricing.

Turn off idle shutdown or change the default inactivity time period

Idle shutdown is enabled and set to shut down your instance after180 inactive minutes by default.

You can change these settings when youcreate an instance.

To turn off idle shutdown or to change the inactivity time periodon an existing instance:

  1. In the Google Cloud console, go to theInstances page.

    Go to Instances

  2. Click the instance name.

  3. On theInstance details page, click theSoftware and security tab.

  4. In theModify software and security configuration section:

    • To turn off idle shutdown, clear the check mark next toEnable Idle Shutdown.

    • To change the inactivity time period,inTime of inactivity before shutdown (Minutes), changethe number to the number of minutes of inactivity that you want.In the Google Cloud console, this settingcan be set to any integer value from 10 to 1440.

  5. ClickSubmit.

Manage idle shutdown using the gcloud CLI

Idle shutdown for Vertex AI Workbench instances is managed usingthemetadata flag. To enable idle shutdown,create an instance with theidle-timeout-seconds key in the metadatawith the value set to the number of seconds.

gcloudworkbenchinstancescreateINSTANCE_NAME--metadata=idle-timeout-seconds=86400

To change the idle shutdown time period, update the value for theidle-timeout-seconds key in the metadata.

gcloudworkbenchinstancesupdateINSTANCE_NAME--metadata=idle-timeout-seconds=43200

To turn off idle shutdown, use the following command:

gcloudworkbenchinstancesupdateINSTANCE_NAME--metadata=idle-timeout-seconds=

Manage idle shutdown on Terraform

Idle shutdown for workbench instances on Terraform is managed using theidle-timeout-seconds key in the metadata field.

To learn how to apply or remove a Terraform configuration, seeBasic Terraform commands.

resource "google_workbench_instance" "default" {  name     = "workbench-instance-example"  location = "us-central1-a"  gce_setup {    machine_type = "n1-standard-1"    vm_image {      project = "cloud-notebooks-managed"      family  = "workbench-instances"    }    metadata = {      idle-timeout-seconds = "10800"    }  }}

How idle shutdown works

Your instance shuts down when there is no kernel activity for thespecified time period. For example, running a cell or new output printingto a notebook is activity that resets the idle shutdown timer. CPU usagedoesn't reset the idle shutdown timer.

If you leave JupyterLab open and you don't interact with the window,the instance will shut down after the idle shutdown period.

By default, idle shutdown looks for activity in kernels running inthe following addresses of the instance:

  • 127.0.0.1:8080/api/sessions
  • 127.0.0.1:8080/api/terminals
  • 127.0.0.1:8080/api/kernels
Warning: By default Jupyter runs on port 8080. If this port is changed, idle shutdown won't detect any kernel activity.

Scheduled executions run while instance is shut down

If you have scheduled an execution of a notebook filein a Vertex AI Workbench instance that is shut down,the execution still runs on schedule.

What's next

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

Last updated 2026-02-19 UTC.