What happens to data disk attached to VM/VMSS when VM/VMSS configured with availability zone fails?

Question

What happens to the data disk when a VM configured with Availability zones failed in any one zone? Does it automatically attach to vm once the vm is spinned up in another zone? Similarly what happens to the data disk when an instance running under vmss(with 1 fixed instance) fails? Does it automatically re-attach when the new instance come up? I did not find any direct answer to this realistic situation. Can I request to clarify me?

Answer 1

Thanks for posting your question in the Microsoft Q&A forum.

When a VM configured with Availability Zones fails in one zone, the data disk does not automatically attach to a new VM in another zone. You would need to manually attach the data disk to the new VM. This is because the data disk is tied to the specific zone where the original VM was running.

https://learn.microsoft.com/en-us/azure/virtual-machines/availability#azure-storage-redundancy

** Please don't forget to close up the thread here by upvoting and accept it as an answer if it is helpful **

Answer 2

Hi @Jeyasingh, Vijayalingam ,

in addition to @Vahid Ghafarpour :

In case of a full outage of an Azure region zone this might affect the data disk as well if the disk isn't from type zone-redundant (ZRS).

For instance:

VM deployed in Azure region zone 1 and data disk (managed disk in zone 1)

In case of a full outage of zone 1 in the Azure region the data disk isn't available in any other zone in the Azure region. You can't attach the data disk to a VM in zone 2 or zone 3.

To get the full advantage of zone redundancy in your scenario you need a data disk with type zone-redundant storage.

In this case the data disk will be synchronously replicated in all zones in the Azure region.

Source: Zone-redundant storage for managed disks

---

(If the reply was helpful please don't forget to upvote and/or accept as answer, thank you)

Regards

Andreas Baumgarten

Answer 3

Hi @Jeyasingh, Vijayalingam ,

I don't know the details of your HA requirements. Also I don't know what kind of service/application is running in the VM(s).

Depending on your HA requirements, a ZRS service might be a good option. If you go for an geo-redundant approach you might run in latency issues in the case of HA failover. This is beside of higher costs and maintenance for HA.

Maybe this link is helpful: Create a Virtual Machine Scale Set that uses Availability Zones

---

(If the reply was helpful please don't forget to upvote and/or accept as answer, thank you)

Regards

Andreas Baumgarten