Route · Zadara → Azure

Migrate VMs from Zadara to Azure

Z-Move, the agentless VM migration platform, adapts the system disk of your Zadara VM for Azure and uploads each disk straight to Azure as a managed disk, sending only the parts that hold data. You then create the VM from that disk in the Azure portal. Nothing is installed in your VMs, and they keep running during the copy.

At a glance

The route, measured

Zadara → Azure

In production
How it runs

Direct upload. Only the parts of each disk that hold data are sent.

You get

An Azure managed disk (Standard SSD). You create the VM from it in the Azure portal.

Operating system

Adapted for Azure, with a check for the Hyper-V drivers.

Measured run
20 GiBdisk size
2.3 GiBactually sent
9m 07stotal time

Validation run, Zadara to Azure East US, upload at 166 MB/s. The upload took 2m 03s with no reconnects: the worker read the whole 20 GiB disk and sent only the 2.3 GiB that held data. Most of the 9m 07s went on starting the worker.

Boot test
Passed: SSH answered on port 22. The test boots a copy of the disk, so the migrated disk stays untouched.
Z-Move session page for a completed Zadara br-ftz-2 to Azure East US migration: 20.0 GiB, transfer 2m 04s at 165.3 MB/s, 9m 07s in total, bottleneck at the source.
A completed Zadara → Azure East US session: 20 GiB in 9m 07s from start to finish, 2m 04s of it transferring. Names from the test environment.

How a Zadara to Azure migration runs

  1. Add the keys of the Zadara region and an Azure service principal (subscription ID, tenant ID, client ID and client secret). They are encrypted and deleted 30 days after the job ends.
  2. Pick the Zadara VM, snapshot or image, the Azure resource group and the name of the disk. The wizard lists your resource groups and checks the name against Azure's rules.
  3. Pre-checks confirm the credentials, that the resource group exists and that the disk name is free before anything starts.
  4. One temporary worker in the source Zadara region snapshots the VM, attaches a volume of each disk and adapts the system disk for Azure (see below).
  5. For each disk, the worker creates an empty managed disk (Standard SSD) in your resource group and uploads the volume to it in parallel, skipping the parts that hold no data. The data goes straight from Zadara to Azure over TLS.
  6. The worker is deleted, and the session lists every resource it created. In the Azure portal, open the disk and create the VM from it.

Adapted for Azure

Before the upload, Z-Move changes the Linux system disk so that it can start on Azure:

  • cloud-init gets the Azure data source first in its list, so it recognises Azure.
  • Network: DHCP on any Ethernet card, because the old configuration is tied to the Zadara network card.
  • Drivers: the worker checks that the kernel has the Hyper-V drivers Azure needs. Without them the disk will not boot on Azure, and the session log warns you.
  • The Hyper-V generation of the disk (1 or 2) is set from its partition table.

What stays as it is

A Windows disk is uploaded as it is, since Windows already ships the Hyper-V drivers. Data disks are uploaded as separate managed disks, named after the system disk.

The disk keeps the system as it was, so it does not depend on the Azure agent or on Azure's provisioning to start, and it works for any operating system.

Before you switch

From the job page, the boot test starts a VM from a copy of the managed disk, in a network of its own, and checks for up to 8 minutes that it answers on SSH, RDP, HTTPS or HTTP, or that the console shows it booted. The copy, the VM and the network are removed at the end, so the migrated disk stays as it was delivered.

In a measured run (Ubuntu, Zadara to Azure East US, 20 GiB), the boot test passed with SSH on port 22. The whole test, cleanup included, took 2m 55s.

Also check these:

  • Size, network and network security group: you choose them in the Azure portal when you create the VM.
  • Security groups and floating IPs of the Zadara region do not move: recreate on Azure what the VM needs.
  • IP addresses change: update DNS records, allow-lists and anything else that points at the old addresses.
  • Only the disks attached to the VM when the job starts are uploaded.

Downtime and rollback

Z-Move copies a snapshot, so the VM keeps running and is never stopped or changed. The copy is crash-consistent, like a server after a power cut. For a database or a busy application, stop the application or the VM before you launch the job: the downtime is then the copy time of that VM.

If an attempt fails, the Azure disks it created are deleted, and the next attempt starts again from the snapshots kept in Zadara. Rolling back means carrying on with the Zadara VM, which stays untouched until you decide to retire it.

What it costs

Z-Move bills US$0.05 per GiB written to the destination: the data, not the size of the disk. A retry does not pay again for disks already copied.

On your Azure bill: the managed disks, from the moment they are created. The temporary worker runs in your Zadara region.

Security

The worker carries no secrets of the platform. It starts with only a token valid for its own job, receives the credentials it needs over TLS while the job runs, keeps them in memory, and accepts no connections from the internet. It writes to each managed disk through a temporary upload link, which it revokes at the end. Credentials are encrypted at rest and never written to logs.

On this route the data goes straight from Zadara to Azure and does not pass through the Z-Move servers.

See the security overview, the DPA and the subprocessors.

FAQ

Zadara to Azure: questions

Something else? Write to hello [at] zmove.click

Do I need to install anything in my Zadara VMs?

No. A temporary worker in the Zadara region reads a snapshot of each disk. Nothing is installed in the VM, and the VM is never stopped or changed.

Why do I get a disk and not a VM?

On Azure, Z-Move produces a managed disk (Standard SSD) in the resource group you choose. You create the VM from it in the Azure portal, where you also pick its size and network.

Is the boot tested on Azure?

Yes. The boot test starts a VM from a copy of the managed disk and checks that it answers on the network or that the console shows it booted; the migrated disk itself is never started. The worker also checks that Linux has the Hyper-V drivers Azure needs, and warns you if they are missing.

Does it work with Windows?

A Windows disk is uploaded as it is, since Windows already ships the Hyper-V drivers that Azure uses. Windows VMs have already been migrated on the Azure routes.

What does the Azure service principal need?

Read access to the subscription and its resource groups, and the right to create, write to, export and delete managed disks in the target resource group. The tightest option is a custom role on that resource group; the simpler one is the built-in Contributor role on that resource group only, never on the whole subscription.

What if the upload fails?

The Azure disks of that attempt are deleted, and the snapshots in Zadara stay for the next attempt, which you start from the same screen. A disk with the chosen name is only replaced if it is an unfinished upload of Z-Move; any other disk is left alone, and the job stops before touching anything.

How long does it take?

It depends mostly on how much of each disk holds data. In a measured run, a 20 GiB disk with 2.3 GiB of data took 9m 07s from start to finish, most of it starting the worker; the upload itself took 2m 03s at 166 MB/s.

Demo

See a full run, from wizard to boot test.

Book a demo and follow a complete migration: wizard, pre-checks, live session, boot test and cleanup.

Or write to hello [at] zmove.click

We only use these details to reply to you.