Live
GitHub secret scanning adds SpaceX partner and refines false‑positive filtersOpen‑Source Nemotron LLMs: What Engineers Need to Know for Self‑Hosted DeploymentRunning Stateful AI Agents on GKE with Agent SubstrateDocker Hardened Images now pull curated packages from ActiveStateBigQuery’s AI.AGG brings native semantic aggregation to petabyte‑scale queriesFull‑Stack Memory Codesign Reduces AI Inference Footprint and Boosts ThroughputDocusign Model Context Protocol hits GA: practical impact on AI‑agent integration and securityA Kubernetes‑Native Pattern for Open‑Weight Model Serving, Fine‑Tuning, and Secure Agent ExecutionGitHub secret scanning adds SpaceX partner and refines false‑positive filtersOpen‑Source Nemotron LLMs: What Engineers Need to Know for Self‑Hosted DeploymentRunning Stateful AI Agents on GKE with Agent SubstrateDocker Hardened Images now pull curated packages from ActiveStateBigQuery’s AI.AGG brings native semantic aggregation to petabyte‑scale queriesFull‑Stack Memory Codesign Reduces AI Inference Footprint and Boosts ThroughputDocusign Model Context Protocol hits GA: practical impact on AI‑agent integration and securityA Kubernetes‑Native Pattern for Open‑Weight Model Serving, Fine‑Tuning, and Secure Agent Execution
Google Cloud

Compute Engine API versioning introduces immutable date‑based snapshots for stable and preview releases

AI SummaryPowered by AI

Google Cloud now offers date‑based versioning for the Compute Engine API, adding quarterly stable snapshots and rapid preview releases alongside the existing /v1 endpoint. This lets engineers adopt new defaults and features on their own schedule without risking production breakage.

Google Cloud has added a date‑based versioning layer to the Compute Engine API, keeping the long‑standing /v1 endpoint while publishing immutable snapshots such as 2026-09-01 (stable) and 2026-10-01-preview (preview). The change gives platform teams a predictable upgrade path and lets them test upcoming features without disrupting live workloads.

Compute Engine API versioning model

The new model replaces the previous channel‑based approach (/v1, /beta, /alpha) with two concrete version families:

  • Stable versions (YYYY‑MM‑DD): Published on a quarterly cadence, starting with 2026-09-01. Each snapshot captures a point‑in‑time contract of all generally available Compute Engine features and defaults. After publication, the snapshot does not receive breaking changes and has no defined end‑of‑support date.
  • Preview versions (YYYY‑MM‑DD‑preview): Issued on a faster cadence (e.g., 2026-10-01-preview) and branch from the most recent stable version. They provide early access to upcoming API designs and have a predictable one‑year testing window.
  • Uninterrupted /v1 support: The existing /v1 endpoint remains fully supported and continues to be the default when no date‑based version is specified.

Why the change matters to AI, cloud, DevOps, and security engineers

For teams that automate VM provisioning, GKE cluster management, or AI‑workload orchestration, the guarantee that /v1 will not change eliminates the need for costly, multi‑year migration plans. At the same time, the stable snapshots let engineers opt‑in to newer security defaults, regional resilience settings, and machine‑readable error formats at a controlled pace. Preview versions give early visibility into upcoming API semantics, allowing test pipelines to validate compatibility before a feature graduates to a stable snapshot.

Operational and security implications

Adopting a date‑based version requires explicit version selection in API calls (e.g., https://compute.googleapis.com/compute/v1/2026-09-01/instances). Teams should audit existing automation to ensure the desired version is referenced, or continue using the default /v1 if no change is needed. Because stable snapshots are immutable, they provide a reliable baseline for compliance audits and reproducible environments. Preview versions, while useful for testing, may introduce non‑breaking changes that could affect error handling or default configurations; they should be exercised in isolated test projects before production rollout.

Related CloudNinjas coverage: Google Cloud.

What This Means For Practitioners

Evaluate your current Compute Engine integrations and decide whether to lock to a specific stable snapshot or remain on /v1. For new projects, consider targeting the latest stable version to inherit the most recent defaults without waiting for a manual migration. Set up a short‑lived test harness that exercises the next preview version to surface any incompatibilities early. Finally, incorporate version selection into your CI/CD pipelines so that upgrades become a deliberate, observable step rather than an implicit change.

Originally published atGoogle Cloud Blog