Release notes - Azure Arc enabled data services (Preview)


As a preview feature, the technology presented in this article is subject to Supplemental Terms of Use for Microsoft Azure Previews.

For current limitations, check the known issues in the release notes.

January 2021

New capabilities and features

Azure Data CLI (azdata) version number: 20.3.0. Download at You can install azdata from Install Azure Data CLI (azdata).

Additional updates include:

  • Localized portal available for 17 new languages

  • Minor changes to Kube-native .yaml files

  • New versions of Grafana and Kibana

  • Issues with Python environments when using azdata in notebooks in Azure Data Studio resolved

  • The pg_audit extension is now available for PostgreSQL Hyperscale

  • A backup ID is no longer required when doing a full restore of a PostgreSQL Hyperscale database

  • The status (health state) is reported for each of the PostgreSQL instances that constitute a sever group

    In earlier releases, the status was aggregated at the server group level and not itemized at the PostgreSQL node level.

  • PostgreSQL deployments now honor the volume size parameters indicated in create commands

  • The engine version parameters is now honored when editing a server group

  • The naming convention of the pods for Azure Arc enabled PostgreSQL Hyperscale has changed

    It is now in the form: ServergroupName{c, w}-n. For example, a server group with three nodes, one coordinator node and two worker nodes is represented as:

    • Postgres01c-0 (coordinator node)
    • Postgres01w-0 (worker node)
    • Postgres01w-1 (worker node)

December 2020

New capabilities & features

Azure Data CLI (azdata) version number: 20.2.5. Download at

View endpoints for SQL Managed Instance and PostgreSQL Hyperscale using the Azure Data CLI (azdata) with azdata arc sql endpoint list and azdata arc postgres endpoint list commands.

Edit SQL Managed Instance resource (CPU core and memory) requests and limits using Azure Data Studio.

Azure Arc enabled PostgreSQL Hyperscale now supports point in time restore in addition to full backup restore for both versions 11 and 12 of PostgreSQL. The point in time restore capability allows you to indicate a specific date and time to restore to.

The naming convention of the pods for Azure Arc enabled PostgreSQL Hyperscale has changed. It is now in the form: ServergroupName{r, s}-n. For example, a server group with three nodes, one coordinator node and two worker nodes is represented as:

  • postgres02r-0 (coordinator node)
  • postgres02s-0 (worker node)
  • postgres02s-1 (worker node)

Breaking change

New resource provider

This release introduces an updated resource provider called Microsoft.AzureArcData. Before you can use this feature, you need to register this resource provider.

To register this resource provider:

  1. In the Azure portal, select Subscriptions
  2. Choose your subscription
  3. Under Settings, select Resource providers
  4. Search for Microsoft.AzureArcData and select Register

You can review detailed steps at Azure resource providers and types. This change also removes all the existing Azure resources that you have uploaded to the Azure portal. In order to use the resource provider, you need to update the data controller and use the latest azdata CLI.

Platform release notes

Direct connectivity mode

This release introduces direct connectivity mode. Direct connectivity mode enables the data controller to automatically upload the usage information to Azure. As part of the usage upload, the Arc data controller resource is automatically created in the portal, if it is not already created via azdata upload.

You can specify direct connectivity when you create the data controller. The following example creates a data controller with azdata arc dc create named arc using direct connectivity mode (connectivity-mode direct). Before you run the example, replace <subscription id> with your subscription ID.

azdata arc dc create --profile-name azure-arc-aks-hci --namespace arc --name arc --subscription <subscription id> --resource-group my-resource-group --location eastus --connectivity-mode direct

Known issues

  • On Azure Kubernetes Service (AKS), Kubernetes version 1.19.x is not supported.

  • On Kubernetes 1.19 containerd is not supported.

  • The data controller resource in Azure is currently an Azure resource. Any updates such as delete is not propagated back to the kubernetes cluster.

  • Instance names can't be greater than 13 characters

  • No in-place upgrade for the Azure Arc data controller or database instances.

  • Arc enabled data services container images are not signed. You may need to configure your Kubernetes nodes to allow unsigned container images to be pulled. For example, if you are using Docker as the container runtime, you can set the DOCKER_CONTENT_TRUST=0 environment variable and restart. Other container runtimes have similar options such as in OpenShift.

  • Cannot create Azure Arc enabled SQL Managed instances or PostgreSQL Hyperscale server groups from the Azure portal.

  • For now, if you are using NFS, you need to set allowRunAsRoot to true in your deployment profile file before creating the Azure Arc data controller.

  • SQL and PostgreSQL login authentication only. No support for Azure Active Directory or Active Directory.

  • Creating a data controller on OpenShift requires relaxed security constraints. See documentation for details.

  • If you are using Azure Kubernetes Service (AKS) Engine on Azure Stack Hub with Azure Arc data controller and database instances, upgrading to a newer Kubernetes version is not supported. Uninstall Azure Arc data controller and all the database instances before upgrading the Kubernetes cluster.

  • AKS clusters that span multiple availability zones are not currently supported for Azure Arc enabled data services. To avoid this issue, when you create the AKS cluster in Azure portal, if you select a region where zones are available, clear all the zones from the selection control. See the following image:

    Clear the checkboxes for each zone to specify none.


  • Azure Arc enabled PostgreSQL Hyperscale returns an inaccurate error message when it cannot restore to the relative point in time you indicate. For example, if you specified a point in time to restore that is older than what your backups contain, the restore will fail with an error message like: ERROR: (404). Reason: Not found. HTTP response body: {"code":404, "internalStatus":"NOT_FOUND", "reason":"Failed to restore backup for server...} When this happens, restart the command after indicating a point in time that is within the range of dates for which you have backups. You will determine this range by listing your backups and by looking at the dates at which they were taken.
  • Point in time restore is supported only across server groups. The target server of a point in time restore operation cannot be the server from which you took the backup. It has to be a different server group. However, full restore is supported to the same server group.
  • A backup-id is required when doing a full restore. By default, if you are not indicating a backup-id the latest backup will be used. This does not work in this release.

October 2020

Azure Data CLI (azdata) version number: 20.2.3. Download at

Breaking changes

This release introduces the following breaking changes:

  • In the PostgreSQL custom resource definition (CRD), the term shards is renamed to workers. This term (workers) matches the command-line parameter name.

  • azdata arc postgres server delete prompts for confirmation before deleting a postgres instance. Use --force to skip prompt.

Additional changes

  • A new optional parameter was added to azdata arc postgres server create called --volume-claim mounts. The value is a comma-separated list of volume claim mounts. A volume claim mount is a pair of volume type and PVC name. The only volume type currently supported is backup. In PostgreSQL, when volume type is backup, the PVC is mounted to /mnt/db-backups. This enables sharing backups between PostgresSQL instances so that the backup of one PostgresSQL instance can be restored in another instance.

  • New short names for PostgresSQL custom resource definitions:

    • pg11
    • pg12
  • Telemetry upload provides user with either:

    • Number of points uploaded to Azure or
    • If no data has been loaded to Azure, a prompt to try it again.
  • azdata arc dc debug copy-logs now also reads from /var/opt/controller/log folder and collects PostgreSQL engine logs on Linux.

  • Display a working indicator during creating and restoring backup with PostgreSQL Hyperscale.

  • azdata arc postrgres backup list now includes backup size information.

  • SQL Managed Instance admin name property was added to right column of overview blade in the Azure portal.

  • Azure Data Studio supports configuring number of worker nodes, vCore, and memory settings for PostgreSQL Hyperscale.

  • Preview supports backup/restore for Postgres version 11 and 12.

September 2020

Azure Arc enabled data services is released for public preview. Arc enabled data services allow you to manage data services anywhere.

  • SQL Managed Instance
  • PostgreSQL Hyperscale

For instructions see What are Azure Arc enabled data services?

Next steps

Just want to try things out?
Get started quickly with Azure Arc Jumpstart on AKS, AWS Elastic Kubernetes Service (EKS), Google Cloud Kubernetes Engine (GKE) or in an Azure VM.