October 2026 Summaries

Pulumi Cloud ESC Approvals now supports concurrent change requests, allowing multiple users to propose independent changes to the same approved environment without waiting for existing requests to be reviewed or applied. Users create drafts as before and can view all requests through the environment’s version dropdown or Approvals tab. If another request is applied first, affected drafts can be updated to the latest revision through a three-way merge, with the console identifying any conflicts for users to resolve by editing or recreating the request. The capability is available immediately for approval-enabled environments on Pulumi Cloud Pro and Enterprise editions.

Pulumi has released version 10 of its Google Cloud provider, based on version 8 of Terraform’s Google provider, introducing breaking changes documented in a comprehensive migration guide. Users can upgrade from v9 manually or through Pulumi’s provider-upgrade skill, available in Pulumi Neo and several coding agents, which scans projects for affected resources, identifies potential replacements or deletions, updates dependencies and code, and runs previews while leaving deployment approval to the user. Version 9 will receive security updates for up to 12 months or until version 11 is released, while new features, bug fixes, and upstream updates will be limited to version 10. New capabilities include support for provisioning Cloud TPU v6e (Trillium) slices and resources for Google’s Gemini Enterprise Agent Platform, including Vertex AI RAG Engine corpora for grounding Gemini applications with custom documents, embedding models, and vector stores.

A team running a Kubernetes-based microservices application uses Pulumi to create an isolated, ephemeral environment for every pull request, allowing reviewers to test changes before merging. Each preview uses a dedicated Pulumi stack and Kubernetes namespace but shares a common infrastructure program with Dev, Stage, and Prod; custom components provision namespace-level resources such as registry credentials and service-level Deployment, Service, and Ingress resources. CI creates a stack and deploys a preview URL when a PR opens, updates the same stack with new container images for subsequent commits, and removes resources after closure, with a configurable grace period. Long-lived shared infrastructure remains outside preview stacks to prevent accidental deletion. At scale, the team found it necessary to serialize updates to individual stacks to avoid concurrent-operation conflicts and to run scheduled reconciliation jobs that retry missed cleanup, address resource drift with Pulumi refresh, and verify deletion. The process typically makes a preview environment reachable in about four minutes while maintaining consistent infrastructure definitions across temporary and permanent environments.