This article answers common questions about Anyscale on Azure. For product-wide questions and self-serve troubleshooting, see the Anyscale on Azure knowledge base.
About the service
What is Anyscale on Azure?
Anyscale on Azure is a managed platform for running distributed Python workloads on Ray. It deploys onto your Azure Kubernetes Service (AKS) cluster as an Azure Native Integration. For details, see What is Anyscale on Azure?.
How is Anyscale on Azure different from the Anyscale SaaS offering?
Anyscale on Azure splits responsibilities across two planes. Anyscale hosts the control plane in its own Azure tenant for scheduling, monitoring, and the Anyscale console. The data plane runs in your Azure subscription on your AKS cluster, so your workloads, container images, and data stay in your tenant. For a full breakdown, see the Architecture overview.
How much does Anyscale on Azure cost?
For current pricing, see Anyscale on Azure pricing. You also pay for the Azure resources that Anyscale on Azure deploys in your subscription, such as the AKS cluster and Azure Container Registry. For the full list, see the Architecture overview.
Deployment and regions
Which Azure regions support Anyscale on Azure?
Anyscale on Azure is available in a limited set of regions. For the current list, see Supported regions.
What Azure resources are deployed by Anyscale on Azure in my subscription?
Deployment creates and uses resources in your Azure subscription, including an AKS cluster, a storage account, an Azure Container Registry, a user-assigned managed identity with federated credentials, and an Azure Load Balancer. For the full picture, see the Architecture overview and Identity and access.
Why does creating an Anyscale cloud fail with a Private Marketplace error?
If your Azure tenant uses Azure Private Marketplace, creating an Anyscale cloud through the Azure portal can fail and return the following error:
The plan "Anyscale on Azure" is not approved in the Private Marketplace.
Azure Private Marketplace blocks the Anyscale offer until a Marketplace administrator approves it for your tenant. To deploy without going through the blocked Marketplace offer, use one of the Terraform examples. Each one stands up the AKS cluster and registers the Anyscale cloud directly:
- Anyscale on AKS reference deployment: The recommended flow for most AKS users.
- Anyscale on AKS Automatic: The same deployment on AKS Automatic, which manages node pools, ingress, and GPU drivers for you.
In what order should I delete Anyscale on Azure resources?
Delete resources in the reverse order that you created them. Terminate running jobs, workspaces, and services, then delete the Anyscale cloud resources, then the Anyscale cloud, then the AKS cluster, and then the resource group. For the steps, see Clean up resources.
The Anyscale operator runs in your AKS cluster and drives workload termination. If you delete the AKS cluster or its resource group first, workloads can't terminate, and the Anyscale console can keep showing jobs, services, or workspaces that no longer exist. If this happens, open a support request in the Azure portal and select Anyscale on Azure as the service. For steps, see Support model.
Identity, security, and data
How do users sign in to Anyscale on Azure?
Users sign in with their existing Microsoft Entra ID credentials through single sign-on. You don't configure a separate identity provider. For details, see Identity and access.
Where does my data and code reside?
Your Ray workloads, container images, and datasets stay in your Azure subscription on your AKS cluster. The Anyscale-hosted control plane handles scheduling and management metadata only. For details, see the Architecture overview.
Support and more information
How do I get support for Anyscale on Azure?
Open a support request through the Azure portal. Azure Support handles all Anyscale on Azure cases and routes them appropriately. For details and current scope, see the Support model.
Where can I find more FAQs and troubleshooting?
For product-wide questions and self-serve troubleshooting, see the Anyscale on Azure knowledge base and the broader Anyscale documentation.