TechPrufer
Get started
Temporal
Temporal logoTemporal · Feature trail

Worker Versioning

5 posts · Mar 2025 – Sep 2026

What this trail covers

Worker Versioning allows for safe deployment of new code changes to Workflows without breaking active executions. This includes concepts like Worker Deployments, Worker Deployment Versions, Build IDs, Pinned and Auto-upgrade Workflows, Current Versions, and ramping versions. The Temporal CLI provides commands for managing and rolling out updates, while the Worker Controller automates these processes in Kubernetes. The Worker Controller now supports autoscaling of Worker versions by allowing the use of new Worker Versioning APIs and a Deployments abstraction for pinning Workflows, ramping traffic, and testing new versions directly.

All Temporal trails
  1. Automated Worker Versioning with GitHub ActionsThis post details the integration of Temporal's Worker Controller with GitHub Actions to automate the Worker Versioning process. It explains how to set up a CI/CD pipeline that triggers on code pushes, builds a Docker image for a new Worker Version, and applies a WorkerDeployment resource. The Worker Controller then manages the creation of versioned Kubernetes Deployments and Worker pods, enabling automated traffic ramping and graceful sunsetting of old worker versions. The post provides a practical example with a GitHub repository and a step-by-step guide to demonstrate the automated deployment in action.
  2. Safe deployments with Temporal Worker Versioning on KubernetesThis post details the practical application of Temporal Worker Versioning on Kubernetes. It demonstrates how to set up a research agent Workflow, intentionally break it with a naive deployment causing non-determinism errors, and then resolve this by installing and using the Temporal Worker Controller. The post explains the concept of "rainbow deployments" where multiple worker versions coexist on the same task queue, with the Temporal Server routing tasks based on worker build IDs. It covers the use of `TemporalWorkerDeployment` resources to manage worker lifecycles, progressive rollouts, and the handling of long-running and idle Workflows during version transitions using `Continue-as-New` and version pinning.
  3. Announcing GA for Worker Versioning and Public Preview for Upgrade on Continue-as-NewAnnounces General Availability for Worker Versioning and Public Preview for Upgrade on Continue-as-New. Worker Versioning allows pinning Workflows to a specific worker version or auto-upgrading them. Upgrade on Continue-as-New enables long-running Workflows that use Continue-as-New to transition to new worker versions at these checkpoints, eliminating the need for manual patching within workflow code. This is achieved by checking `target_worker_deployment_version_changed` and using `ContinueAsNewErrorOptions` with `InitialVersioningBehavior: workflow.ContinueAsNewVersioningBehaviorAutoUpgrade`.
  4. Safe, versioned Worker Deployments on Kubernetes: Now with autoscaling!Introduces `WorkerResourceTemplate`, a new custom resource for the Temporal Worker Controller that allows users to define Kubernetes resources (like HPAs and PDBs) to be automatically created and managed for each versioned Worker Deployment. This enables autoscaling of individual worker versions based on metrics such as CPU utilization, slot utilization, and task queue backlog count. The feature leverages Kubernetes patterns and existing autoscaling tooling, providing security guardrails through a validating webhook to prevent privilege escalation and block banned resource kinds.
  5. Announcing Worker Versioning Public Preview: Pin Workflows to a single code versionIntroduces Worker Versioning, a feature allowing Workflows to be pinned to a single code version (Pinned Workflows) or automatically upgrade (Auto-upgrade Workflows). This is enabled by a built-in task router that substitutes for traditional load balancers and first-class support for rainbow deployments. A Temporal Worker Controller for Kubernetes is also provided as a drop-in solution. The post details three deployment strategies: rolling, blue-green, and rainbow, highlighting the benefits of rainbow deploys for long-running, multi-process Workflows. It also outlines the design philosophy (ease of adoption, reliability, deployment safety, flexibility) and project status (public preview) with a roadmap towards General Availability (GA) in Q4 2025.
  7. Announcing KEDA-based auto-scaling for Temporal WorkersIntroduces the integration of KEDA (Kubernetes-based Event-Driven Autoscaling) with Temporal Workers, specifically the Temporal Worker scaler available in KEDA v2.17.0. This enables dynamic scaling of Temporal workers based on Task Queue backlog size, offering cost efficiency, high throughput, and operational simplicity. A demo application (Order Management System) is provided to showcase this functionality.
  8. Replay 2025: New Temporal Capabilities for Faster DevelopmentIntroduced new Worker Versioning APIs and a Deployments abstraction (Pre-release) to enable "pinning" Workflows to specific Deployment Versions for safer rollouts, ramping traffic between versions, and direct testing of new Deployment Versions before full rollout. These can be managed via APIs or a Kubernetes controller.