TechPrufer
Sign inGet started

Blogs›Temporal›Worker Versioning

Worker Versioning

Worker Versioning

4
posts
2025–2026

Worker Versioning is now Generally Available, allowing Workflows to be pinned to specific code versions or automatically upgraded. Upgrade on Continue-as-New is now in Public Preview, enabling long-running Workflows that use Continue-as-New to seamlessly upgrade to new worker versions at their checkpoint.

2026

Safe deployments with Temporal Worker Versioning on Kubernetes

4/22/2026

This post details the practical implementation and benefits of Temporal Worker Versioning on Kubernetes. It explains how to use the Temporal Worker Controller to manage multiple worker versions simultaneously on the same task queue, preventing non-determinism errors during deployments. The post walks through a hands-on example of building a research agent workflow, intentionally breaking it with a naive deployment, and then fixing it using Worker Versioning. It covers the installation of the Worker Controller, deployment of v1 and v2 workers, and the handling of long-running and sleeping workflows during updates, including the use of Continue-as-New and the caveat of idle workflows not noticing version changes.

Announcing GA for Worker Versioning and Public Preview for Upgrade on Continue-as-New

3/30/2026

Introduces 'Upgrade on Continue-as-New' as a Public Preview feature, which allows Workflows to transition to new worker versions at Continue-as-New points. This complements the GA of Worker Versioning by providing a strategy for upgrading very long-running Workflows without code patching.

Safe, versioned Worker Deployments on Kubernetes: Now with autoscaling!

3/30/2026

Introduced the `WorkerResourceTemplate` custom resource for the Temporal Worker Controller, enabling the automatic attachment and lifecycle management of Kubernetes resources (e.g., HPAs, PDBs) to versioned Temporal Worker Deployments. This allows for independent autoscaling of each worker version based on metrics like CPU utilization or Task Queue backlog, and ensures automatic cleanup of associated resources when a version is retired. Enhanced metric reporting by adding worker version labels to the `approximate_backlog_count` metric.

2025

Announcing Worker Versioning Public Preview: Pin Workflows to a single code version

9/24/2025

Introduces Worker Versioning with two behaviors: PINNED (Workflows stay on a single version) and AUTO_UPGRADE (Workflows automatically hop to new versions). Details three deployment strategies: rolling, blue-green, and rainbow, with a focus on how Worker Versioning enables safer rainbow deployments. Discusses design goals (ease of adoption, reliability, deployment safety, flexibility) and provides links to documentation, a Kubernetes Worker Controller, and a course. Outlines project status (public preview) and roadmap, including plans for GA, Pinned-until-Continue-as-New annotation, replay testing, and auto-scaling for the Worker Controller.