TechPrufer
Get started
Temporal
Temporal logoTemporal · Feature trail

Temporal Worker Deployment on AWS ECS

1 post · Dec 2025 – Apr 2026

What this trail covers

Temporal Workers can be deployed to Kubernetes, with specific guidance on containerization, Kubernetes Deployment manifests, and the use of ConfigMaps and Secrets. The post emphasizes leveraging standard Kubernetes patterns, the Temporal Worker Controller for advanced version management, and Worker Versioning for safe deployments. It also details autoscaling strategies for Temporal Workers on Kubernetes, highlighting the inadequacy of standard CPU/memory metrics and recommending Task Queue Backlog, Schedule-to-Start latency, and Worker Task slots as more effective signals. KEDA-based autoscaling is presented as a streamlined solution. For the Temporal Service, official Helm charts are provided, with recommendations to disable bundled dependencies and use external ones for production. Planning capacity and tuning for production involves iterative load testing and avoiding CPU throttling. The post also touches on the need for graceful shutdown of workers.

All Temporal trails
  1. Deploying Temporal Workers to Amazon ECSThis post details the configuration of Temporal Workers for deployment on Amazon ECS with Fargate. It covers essential aspects such as the Worker container setup (Dockerfile, dependency caching), implementing health checks via a lightweight HTTP endpoint, and handling graceful shutdowns using SIGTERM signals. Infrastructure considerations include ECS task definitions (no port mappings, secrets via SSM Parameter Store), IAM roles (Execution vs. Task roles), and ECS service configuration emphasizing Fargate Spot for cost savings. Scaling strategies focus on ECS Application Auto Scaling with low CPU targets and asymmetric cooldowns. Observability is addressed through CloudWatch logs, Container Insights, and Temporal Cloud metrics. The post also provides guidance on choosing between ECS and EKS based on team expertise, fleet size, and specific operational needs.
  2. Tips for running Temporal on KubernetesThis post provides practical tips for deploying and operating Temporal on Kubernetes. It details containerizing Temporal Workers, creating Kubernetes Deployment manifests, and using ConfigMaps/Secrets. It recommends leveraging standard Kubernetes patterns, the Temporal Worker Controller, and Worker Versioning for advanced deployment management. The post also elaborates on autoscaling Temporal Workers on Kubernetes, explaining why standard CPU/memory metrics are insufficient and proposing Task Queue Backlog, Schedule-to-Start latency, and Worker Task slots as better signals. It also mentions KEDA-based autoscaling. For the Temporal Service, it advises using official Helm charts, disabling bundled dependencies, and using external ones for production. Capacity planning and tuning for production are discussed, emphasizing iterative load testing and avoiding CPU throttling. The importance of graceful shutdown for workers is also highlighted.