cloud-init support for virtual machines in Azure


This article references CentOS, a Linux distribution that is nearing End Of Life (EOL) status. Please consider your use and plan accordingly.

Applies to: ✔️ Linux VMs ✔️ Flexible scale sets

This article explains the support that exists for cloud-init to configure a virtual machine (VM) or Virtual Machine Scale Sets at provisioning time in Azure. These cloud-init configurations are run on first boot once the resources have been provisioned by Azure.

VM Provisioning is the process where the Azure will pass down your VM Create parameter values, such as hostname, username, and password, and make them available to the VM as it boots up. A 'provisioning agent' will consume those values, configure the VM, and report back when completed.

Azure supports two provisioning agents cloud-init, and the Azure Linux Agent (WALA).

cloud-init overview

cloud-init is a widely used approach to customize a Linux VM as it boots for the first time. You can use cloud-init to install packages and write files, or to configure users and security. Because cloud-init is called during the initial boot process, there are no additional steps or required agents to apply your configuration. For more information on how to properly format your #cloud-config files or other inputs, see the cloud-init documentation site. #cloud-config files are text files encoded in base64.

cloud-init also works across distributions. For example, you don't use apt-get install or yum install to install a package. Instead you can define a list of packages to install. cloud-init automatically uses the native package management tool for the distro you select.

We're actively working with our endorsed Linux distro partners in order to have cloud-init enabled images available in the Azure Marketplace. These images will make your cloud-init deployments and configurations work seamlessly with VMs and virtual machine scale sets. Initially we collaborate with the endorsed Linux distro partners and upstream to ensure cloud-init functions with the OS on Azure, then the packages are updated and made publicly available in the distro package repositories.

There are two stages to making cloud-init available to the supported Linux distributions on Azure, package support, and then image support:

  • 'cloud-init package support on Azure' documents, which cloud-init packages onwards are supported or in preview, so you can use these packages with the OS in a custom image.
  • 'image cloud-init ready' documents if the image is already configured to use cloud-init.


Publisher / Version Offer SKU Version image cloud-init ready cloud-init package support on Azure
Canonical 24.04 UbuntuServer 22.04-LTS latest yes yes
Canonical 22.04 UbuntuServer 22.04-LTS latest yes yes
Canonical 20.04 UbuntuServer 20.04-LTS latest yes yes


Publisher / Version Offer SKU Version image cloud-init ready cloud-init package support on Azure
RedHat 7 RHEL 7.7, 7.8, 7_9 latest yes yes
RedHat 8 RHEL 8.1, 8.2, 8_3, 8_4 latest yes yes
RedHat 9 RHEL 9_0, 9_1 latest yes yes
  • All other RedHat SKUs starting from RHEL 7 (version 7.7) and RHEL 8 (version 8.1) including both Gen1 and Gen2 images are provisioned using cloud-init. Cloud-init is not supported on RHEL 6.


Publisher / Version Offer SKU Version image cloud-init ready cloud-init package support on Azure
OpenLogic 7 CentOS 7.7, 7.8, 7.9 latest yes yes
OpenLogic 8 CentOS 8.1, 8.2, 8.3 latest yes yes
  • All other CentOS SKUs starting from CentOS 7 (version 7.7) and CentOS 8 (version 8.1) including both Gen1 and Gen2 images are provisioned using cloud-init. CentOS 6.10, 7.4, 7.5, and 7.6 images don't support cloud-init.


Publisher / Version Offer SKU Version image cloud-init ready cloud-init package support on Azure
Oracle 7 Oracle Linux 77, 78, ol79 latest yes yes
Oracle 8 Oracle Linux 81, ol82, ol83-lvm, ol84-lvm latest yes yes
  • All other Oracle SKUs starting from Oracle 7 (version 7.7) and Oracle 8 (version 8.1) including both Gen1 and Gen2 images are provisioned using cloud-init.


Publisher / Version Offer SKU Version image cloud-init ready cloud-init package support on Azure
SUSE 15 SLES (SUSE Linux Enterprise Server) sp1, sp2, sp3 latest yes yes
SUSE 12 SLES (SUSE Linux Enterprise Server) sp5 latest yes yes


Publisher / Version Offer SKU Version image cloud-init ready cloud-init package support on Azure
debian-10 Debian 10-cloudinit-gen2 Debian:debian-10:10-cloudinit-gen2:0.0.1015 yes yes
debian-10 Debian 10-cloudinit-gen2 Debian:debian-10:10-cloudinit-gen2:0.0.991 yes yes
debian-10 Debian 10-cloudinit-gen2 Debian:debian-10:10-cloudinit-gen2:0.0.999 yes yes

Currently Azure Stack will support the provisioning of cloud-init enabled images.

What is the difference between cloud-init and the Linux Agent (WALA)?

WALA is an Azure platform-specific agent used to provision and configure VMs, and handle Azure extensions.

We're enhancing the task of configuring VMs to use cloud-init instead of the Linux Agent in order to allow existing cloud-init customers to use their current cloud-init scripts, or new customers to take advantage of the rich cloud-init configuration functionality. If you have existing investments in cloud-init scripts for configuring Linux systems, there are no additional settings required to enable cloud-init process them.

cloud-init cannot process Azure extensions, so WALA is still required in the image to process extensions, but will need to have its provisioning code disabled, for endorsed Linux distros images that are being converted to provision by cloud-init, they will have WALA installed, and setup correctly.

When creating a VM, if you don't include the Azure CLI --custom-data switch at provisioning time, cloud-init or WALA takes the minimal VM provisioning parameters required to provision the VM and complete the deployment with the defaults. If you reference the cloud-init configuration with the --custom-data switch, whatever is contained in your custom data will be available to cloud-init when the VM boots.

cloud-init configurations applied to VMs do not have time constraints and will not cause a deployment to fail by timing out. This isn't true for WALA, if you change the WALA defaults to process custom-data, it can't exceed the total VM provisioning time allowance of 40 minutes, if so, the VM Create will fail.

cloud-init VM provisioning without a UDF driver

Beginning with cloud-init 21.2, you can use cloud-init to provision a VM in Azure without a UDF driver. If a UDF driver isn't available in the image, cloud-init uses the metadata that's available in the Azure Instance Metadata Service to provision the VM. This option works only for SSH key and user data. To pass in a password or custom data to a VM during provisioning, you must use a UDF driver.

Deploying a cloud-init enabled Virtual Machine

Deploying a cloud-init enabled virtual machine is as simple as referencing a cloud-init enabled distribution during deployment. Linux distribution maintainers have to choose to enable and integrate cloud-init into their base Azure published images. Once you've confirmed the image you want to deploy is cloud-init enabled, you can use the Azure CLI to deploy the image.

The first step in deploying this image is to create a resource group with the az group create command. An Azure resource group is a logical container into which Azure resources are deployed and managed.

The following example creates a resource group named myResourceGroup in the eastus location.

az group create --name myResourceGroup --location eastus

The next step is to create a file in your current shell, named cloud-init.txt and paste the following configuration. For this example, create the file in the Cloud Shell not on your local machine. You can use any editor of your choice. Enter sensible-editor cloud-init.txt to create the file and see a list of available editors. Use the editor of your choice. Some typical choices are nano, vim, or ed. Make sure that the whole cloud-init file is copied correctly, especially the first line:

# cloud-config
package_upgrade: true
- apache2
# cloud-config
package_upgrade: true
- httpd
# cloud-config
package_upgrade: true
- httpd


cloud-init has multiple input types, cloud-init will use first line of the customData/userData to indicate how it should process the input, for example #cloud-config indicates that the content should be processed as a cloud-init config.

Exit the file and save the file according to the editor. Verify file name on exit.

The final step is to create a VM with the az vm create command.

The following example creates a VM named centos74 and creates SSH keys if they don't already exist in a default key location. To use a specific set of keys, use the --ssh-key-value option. Use the --custom-data parameter to pass in your cloud-init config file. Provide the full path to the cloud-init.txt config if you saved the file outside of your present working directory.

az vm create \
  --resource-group myResourceGroup \
  --name centos74 \
  --image OpenLogic:CentOS-CI:7-CI:latest \
  --custom-data cloud-init.txt \

When the VM has been created, the Azure CLI shows information specific to your deployment. Take note of the publicIpAddress. This address is used to access the VM. It takes some time for the VM to be created, the packages to install, and the app to start. There are background tasks that continue to run after the Azure CLI returns you to the prompt. You can SSH into the VM and use the steps outlined in the Troubleshooting section to view the cloud-init logs.

You can also deploy a cloud-init enabled VM by passing the parameters in ARM template.

Troubleshooting cloud-init

Once the VM has been provisioned, cloud-init will run through all the modules and script defined in --custom-data in order to configure the VM. If you need to troubleshoot any errors or omissions from the configuration, you need to search for the module name (disk_setup or runcmd for example) in the cloud-init log - located in /var/log/cloud-init.log.


Not every module failure results in a fatal cloud-init overall configuration failure. For example, using the runcmd module, if the script fails, cloud-init will still report provisioning succeeded because the runcmd module executed.

For more details of cloud-init logging, see the cloud-init documentation


cloud-init collects usage data and sends it to Microsoft to help improve our products and services. Telemetry is only collected during the provisioning process (first boot of the VM). The data collected helps us investigate provisioning failures and monitor performance and reliability. Data collected doesn't include any identifiers (personal identifiers). Read our privacy statement to learn more. Some examples of telemetry being collected are (this isn't an exhaustive list): OS-related information (cloud-init version, distro version, kernel version), performance metrics of essential VM provisioning actions (time to obtain DHCP lease, time to retrieve metadata necessary to configure the VM, etc.), cloud-init log, and dmesg log.

Telemetry collection is currently enabled for most of our marketplace images that use cloud-init. It is enabled by specifying KVP telemetry reporter for cloud-init. In most Azure Marketplace images, this configuration can be found in the file /etc/cloud/cloud.cfg.d/10-azure-kvp.cfg. Removing this file during image preparation will disable telemetry collection for any VM created from this image.

Sample content of 10-azure-kvp.cfg

    type: log
    type: hyperv

Next steps

Troubleshoot issues with cloud-init.

For cloud-init examples of configuration changes, see the following documents: