Allow idle timeout to be set on Notebook Instances (Vertex AI Workbench)

[oscar-barlow]

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
• Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritize the request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment. If the issue is assigned to the "modular-magician" user, it is either in the process of being autogenerated, or is planned to be autogenerated soon. If the issue is assigned to a user, that user is claiming responsibility for the issue. If the issue is assigned to "hashibot", a community member has claimed the issue already.

Description

Leaving notebook VMs running when they're not being actively used is a waste of money, so GCP offer an 'idle timeout' configuration option in the Vertex Workbench console. It would be really helpful to be able to configure this in Terraform, to save money.

• Vertex AI > Documentation > Vertex AI Workbench > Guides > Idle Shutdown
• Vertex AI Workbench release notes for July 19 2023

New or Affected Resource(s)

• google_notebooks_instance

Potential Terraform Configuration

resource "google_notebooks_instance" "instance" {
  name = "notebooks-instance"
  location = "us-west1-a"
  machine_type = "e2-medium"
  idle_shutdown = 60 # 1 hour in minutes
  vm_image {
    project      = "deeplearning-platform-release"
    image_family = "tf-latest-cpu"
  }
}

b/299312211

[rileykarson]

This may be managed through the unstructuredmetadata field. That's how it looks like it's done in the v2 API, at least, snippet from a REST response:

        "metadata": {
          "idle-timeout-seconds": "10800",
... other entries ...
        },

[mahvar]

@rileykarson Is that already doable via metadata?

[enlightenalpha]

Can confirm that the metadata approach described above worked for me:

"metadata": { "idle-timeout-seconds": "10800" }

[bcreddy-gcp]

@rileykarson Is there some documentation that'll help the users discover this better? Seems to be a very common question. Otherwise this ticket can be closed.

[bcreddy-gcp]

Documentation has been added here: https://cloud.google.com/vertex-ai/docs/workbench/instances/idle-shutdown

[markustoivonen]

Sorry if this is a stupid question but is the correct syntax?

resource "google_workbench_instance" "workbench_instance_test" {
  ...
  metadata = {
    idle-timeout-seconds = 1000
  }

Also, would it be possible to add the information to the resource documentation, considering that this is a new feature, and something that is new at least for those who are migrating from user managed notebooks.

[bcreddy-gcp]

Here's the syntax:

resource "google_workbench_instance" "workbench_instance_test" {
  ...
  metadata = {
    idle-timeout-seconds = "10800"
  }

[sonnysideup]

This worked for me as well. Thanks for the tip.

[codrinsah]

I haven't seen this mentioned yet, but if a notebook's idle timeout was already set e.g.

metadata = {
  idle-timeout-seconds = "10800"
}

In order to disable the idle timeout, you need to set the metadata to an empty string (not null or removing it!) like so:

metadata = {
  idle-timeout-seconds = ""
}

I noticed that simply removing it would not change the setting of the notebook.