Skip to main content
resource · gcp

Vertex AI Notebook Runtime

schedulable
no
category
ai-ml-services

Does ZopNight manage Vertex AI Notebook Runtime?

Colab Enterprise notebook runtimes bill for their provisioned VM and any GPU for as long as the runtime exists, not just while a session is active. ZopNight discovers runtimes via Cloud Asset Inventory and flags long-lived idle ones in recommendations, since a runtime that outlives its session bills like an idle VM.

Rules that fire on Vertex AI Notebook Runtime

no live rules

No active rule family targets Vertex AI Notebook Runtime today. Rules that used to are retired, and retired rules publish no pages and fire no findings. Scheduling and permissions coverage are unaffected.

Browse every live recommendation for this platform →

A notebook runtime is the provisioned compute backing a Colab Enterprise notebook session, billed while it exists. Runtimes that outlive their sessions keep billing like idle VMs.

Colab Enterprise compute that persists after the session

A runtime bills for its VM and any accelerator for the whole time it stays provisioned. Provisioned is not the same as used. Closing the browser tab does not delete the runtime; unless the idle-shutdown policy inherited from its template retires it, the machine sits allocated and metering. That gap between session-over and runtime-gone is the entire cost story of this type, and it is invisible from inside the notebook UI, where everything looks closed.

Idle-runtime detection

ZopNight discovers notebook runtimes via Cloud Asset Inventory and flags long-lived idle runtimes in its recommendations. The type is the Colab Enterprise counterpart to Workbench instances, but without a schedulable stop verb in ZopNight, so the remedies are deleting abandoned runtimes and, upstream, templates that enforce idle shutdown. A runtime’s cost profile is fixed at creation by its template: machine shape, accelerators, and the idle-shutdown timer all inherit from there.

Runtimes that outlive their notebooks

The recurring leaks: runtimes created from templates with idle shutdown disabled, persisting for days after their session ended; GPU runtimes spun up for one experiment and abandoned when the notebook moved to someone else’s runtime instead; and a pileup of per-user runtimes in shared projects, each modest on its own but collectively a standing fleet. Every one of them is a VM bill wearing a notebook costume.

Colab runtimes in the console

Google Cloud console → Vertex AI → Colab Enterprise → Runtimes shows every runtime with its machine type and creation time. Old creation timestamps on supposedly temporary runtimes are the tell: a runtime created weeks ago for a one-day experiment has been billing the whole time, and deleting it costs nothing but the click.

See it fire on your bill.

Connect an account read-only. The first findings land in minutes.

417 rule families across 353 resource types on 22 platforms. Every threshold, metric, and IAM action is documented on these pages before you grant anything.

417 rule families documented
353 resource types covered
read-only default access level
Multi-cloud automation· Production-ready in 30 min· SOC 2 · ISO 27001· 20–60% off the bill, first month· 4 platforms · 1 console· Multi-cloud automation· Production-ready in 30 min· SOC 2 · ISO 27001· 20–60% off the bill, first month· 4 platforms · 1 console·