AzureVM: managing virtual machines in Azure

[This article was first published on Revolutions, and kindly contributed to R-bloggers]. (You can report issue about the content on this page here)
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

This is the next article in my series on AzureR, a family of packages for working with Azure in R. I’ll give a short introduction on how to use AzureVM to manage Azure virtual machines, and in particular Data Science Virtual Machines (DSVMs).

Creating a VM

Creating a VM is as simple as using the create_vm method, which is available as part of the az_subscription and az_resource_group classes.


## using the subscription method
sub <- az_rm$
    new(tenant="{tenant_id}", app="{app_id}", password="{password}")$

myNewVM <- sub$create_vm("myNewVM",
    passkey="fAs30q-2a5vF!Z")  # be sure to choose a strong password!

## using the resource group method
rg <- sub$create_resource_group("myresourcegroup",

myOtherVM <- rg$create_vm("myOtherVM",

Without any other options, this will create a Windows Server 2016 Data Science Virtual Machine, which is pre-installed with several tools useful for analytics: Python, R (and RStudio), Tensorflow, XGBoost, SQL Server, and so on. The size will be a Standard DS3 V2 VM, which has 4 cores, 14GB of memory, 1TB primary disk, and up to 16x28GB data disks.

A feature of creating the VM under a subscription, as opposed to a resource group, is that it creates a new resource group specifically to hold the VM. This simplifies the task of managing and (eventually) deleting the VM considerably. See “Deleting a VM” below.

You can change the specifications for the VM by providing any of the following arguments:

  • os: either “Windows” or “Ubuntu” (for Ubuntu LTS 16.04).
  • size: the VM size. Use the az_subscription$list_vm_sizes() method to see what sizes are available in your region. Note that in Azure, a VM’s size is really a broad-ranging label that encapsulates the number of cores, memory, and disk available.
  • passkey: if creating an Ubuntu VM, you can supply a public key as the passkey argument.
  • userauth_type: set this to “key” if you supply a public key.

Setting these will determine whether you get a Windows or Ubuntu DSVM, the login details, and how powerful the VM is in terms of cores and memory/disk capacity. For example, this will create a Linux NC-series DSVM using public key authentication:

myLinuxVM <- sub$create_vm("myLinuxVM",

The list_vm_sizes() method will show you what the available VM sizes are for your region.

# examine VM sizes available in australiaeast region

Retrieving an existing VM

If you have an existing VM, you can retrieve it with the get_vm() method. As with create_vm(), this is available as part of the az_subscription and az_resource_group classes. The only argument you need to supply is the name of the VM.

## using the subscription method
sub <- az_rm$
    new(tenant="{tenant_id}", app="{app_id}", password="{password}")$

# retrieve the VM we created above
myNewVM <- sub$get_vm("myNewVM")

## and with the resource group method
rg <- sub$get_resource_group("myresourcegroup")

myOtherVM <- rg$get_vm("myOtherVM")

Working with a VM

There are various things you can do with a VM object within R.

To stop (shutdown) a VM, call its stop() method. The deallocate argument sets whether to deallocate its resources as well, with the default being TRUE; you may want to set this to FALSE if you know you will be restarting the VM soon. To restart it, call either the start() or restart() method. The main difference between the two is that restart() will shutdown the VM first if it is currently running.

To sync the object with the resource in Azure, call the sync_vm_status() method. The most common situation where you might want to do this is if you create a VM with the argument wait=FALSE. In this case, rather than waiting for provisioning to complete, the create_vm method will return an incomplete VM object; you then call its sync_vm_status() method to update it with how the provisioning is going in Azure.

To dynamically resize a VM, call the resize() method with the new size. This has an optional deallocate argument controlling whether to stop and deallocate the VM first (which is sometimes necessary for successful resizing).

To run a script in the VM (without manually logging in), call the run_script() method. This will be a PowerShell script if it is a Windows VM, or a bash script if it is Linux. The script is just a character vector.

# simple bash script for executing on a Linux VM
script <-

var=Hello world!

# redirect output to a file so we know whether it ran successfully
echo "$var" > /tmp/helloworld.txt


If you login to the VM after running this script, you should find the file helloworld.txt in the /tmp directory.

Deleting a VM

Simply deleting a virtual machine object in R, eg with rm(), will not do anything to the VM itself in Azure. To delete the VM and its resources, call the object’s delete() method:


AzureVM will prompt you for confirmation that you really want to delete the VM. By default, this will also free up all the individual Azure resources used by the VM, such as its storage, network interface, security group, and so on.

If you created the VM using the create_vm() method of the az_subscription class, the deletion process is very simple: it simply removes the resource group containing the VM. This is possible because the resource group was created as part of deploying the VM. Be aware that any other resources you may have created in this resource group will also be deleted.

Alternatively, you can use the delete_vm() method of the az_subscription and az_resource_group classes. These will retrieve the VM of the given name and then call its delete() method.

GPU enabled VMs

If you are running deep learning workloads, you’ll want to ensure that your VM is GPU-enabled. In Azure, the various NC- and ND-series VMs are designed for these workloads. You can create a GPU-enabled VM by setting the size argument appropriately, for example

myGpuVM <- sub$create_vm(size="Standard_NC12s_v3", ...)

However, the following caveats apply to GPU-enabled VMs: - Not all regions have GPUs available. To check on availability, use the az_subscription$list_vm_sizes() method and provide your region. - Currently, the supply of GPUs is limited. You must apply for a quota increase if you want to deploy a GPU-enabled VM.

VM clusters

You can work with VM clusters (a collection of VMs sharing the same configuration) by using the get_vm_cluster, create_vm_cluster and delete_vm_cluster methods. These are almost identical to get_vm, create_vm and delete_vm with the addition of a clust_size argument that sets the size of the cluster.

sub <- az_rm$
    new(tenant="{tenant_id}", app="{app_id}", password="{password}")$

# create a cluster of 5 Ubuntu VMs
vmCluster <- sub$create_vm_cluster("vmCluster",

Most things that you can do with a single VM, you can also do with a VM cluster. For example, running a script with the run_script() method will run the script on all the VMs in the cluster. Starting, stopping and restarting a cluster similarly carries out the given action on all the VMs.

To leave a comment for the author, please follow the link and comment on their blog: Revolutions. offers daily e-mail updates about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

Never miss an update!
Subscribe to R-bloggers to receive
e-mails with the latest R posts.
(You will not see this message again.)

Click here to close (This popup will not appear again)