Google Cloud ACE Cloud Run Operations Guide

Study Google Cloud ACE Cloud Run Operations: key concepts, common traps, and exam decision cues.

This lesson is about running Cloud Run safely after deployment. Google Cloud expects you to know how revisions, traffic splitting, and scaling controls affect release safety and live behavior.

Revision: Immutable deployed version of a Cloud Run service that can receive traffic independently.

Traffic splitting: Controlled routing of percentages of live traffic across revisions.

Scaling controls: Runtime settings that affect how quickly and how far Cloud Run instances grow or stay warm.

What Google Cloud is really testing here

ACE wants you to separate:

  • deploying a new version from sending traffic to it
  • revision identity from service identity
  • safe rollout from full cutover
  • traffic behavior from runtime scale behavior

Many stems are release-management questions disguised as serverless questions.

Fast release chooser

If the question is mainly about… Strongest first lane
keeping the old version available while testing a new one separate revision
sending only part of production traffic to the new version traffic splitting
rolling back quickly after a bad release move traffic back to the earlier healthy revision
cold-start sensitivity or controlling live capacity behavior scaling controls

Safe rollout mental model

    flowchart LR
	  A["Deploy new revision"] --> B["Keep old revision live"]
	  B --> C["Send small traffic share to new revision"]
	  C --> D["Observe errors and latency"]
	  D --> E["Shift more traffic or roll back"]

That order matters. A new revision existing does not mean it should receive all production traffic immediately.

Revision versus scaling knobs

Question Strongest first answer
canary or gradual rollout traffic split across revisions
instant rollback after a bad release send traffic back to the prior healthy revision
latency spikes from cold starts or burst behavior scaling settings
“which version is live?” check revision and traffic assignment, not just the latest deployment

Common traps

Trap Better reading
“The newest revision should automatically get all traffic.” ACE expects you to think about controlled rollout, not automatic full cutover.
“A revision is just a log entry.” A revision is a real deployable version boundary for routing and rollback.
“Traffic splitting and scaling are the same control.” Traffic splitting controls who gets requests. Scaling controls runtime capacity behavior.
“Rollback means redeploying from scratch.” If the previous revision is healthy, routing traffic back is usually the fastest safe response.

Harder scenario question

A new Cloud Run release is live for 10% of users and starts showing errors. The team wants the fastest safe recovery. Which lane is strongest first?

  • A. Resize the subnet
  • B. Send traffic back to the known-good revision
  • C. Rotate every IAM binding
  • D. Delete the service and recreate it later

Correct answer: B. If the failure is isolated to the new revision, the fastest safe move is to shift traffic back rather than destroy the service or make unrelated infrastructure changes.

Decision order that usually wins

  1. Separate versioned rollout control, fast rollback, and runtime scaling behavior.
  2. If only part of the traffic should hit a new version, think traffic splitting across revisions.
  3. If a new rollout errors under partial exposure, think shift traffic back to the last healthy revision.
  4. If the concern is cold starts or steadier burst handling, think Cloud Run scaling controls.
  5. ACE usually rewards using revision-aware controls before making unrelated infrastructure changes.

Quiz

Loading quiz…
Revised on Sunday, May 10, 2026