Is Windows Server 2019 Failover Cluster Possible Without Shared Storage?

Question

Can two Windows Server 2019 servers, each with their own internal storage, be used in a failover cluster without using external shared storage? The company wants a complete backup of their Domain, SQL and Exchange Servers (three separate servers, not virtual), replicated in another branch office, so if the main office suffers a disaster (fire, flood, earthquake), the branch office could take over. All of the failover cluster topologies I have seen use shared storage. It seems that would lead to a single disaster point bringing down the entire failover system.

Any insight would be appreciated.

Answer 1

Hi. Thank you for your question and reaching out. I’d be more than happy to help you with your query

Yes, Windows Server 2019 Failover Cluster is possible without shared storage. However, it is not recommended to use this configuration in production environments as it may cause issues with data availability and consistency. Additionally, without shared storage, the cluster will not be able to failover and failback without manual intervention. It is recommended that you use shared storage when configuring a failover cluster, as this will ensure that the data is available and consistent across all nodes in the cluster, even when nodes fail.

If the reply was helpful, please don’t forget to upvote or accept as answer, thank you.

Answer 2

Hey,

As mentioned, it is possible to create Failover Cluster without shared storage (physical SAN). You will need to use Software-Defined Storage such as S2D or StarWind VSAN. SDS will create shared storage pool, utilizing local storage of your nodes. Check for more information: https://www.starwindsoftware.com/software-defined-storage

Cheers, 
Alex Bykovskyi 
StarWind Software 
Note: Posts are provided “AS IS” without warranty of any kind, either expressed or implied, including but not limited to the implied warranties of merchantability and/or fitness for a particular purpose.

Answer 3

A failover cluster requires some form of shared storage. Shared storage can come in various solutions, iSCSI, shared SAS, Fibre Channel, and SMB. A specialized version of clustering, Storage Spaces Direct (S2D), can be configured using locally attached storage. Clustered applications always use shared storage. That enables the application to run on any node of the cluster to access the storage needed for the application. Generally, when talking about high availability of shared storage, some method is employed to provide HA at the storage level. SAN systems often provide that as part of their construction - duplicate controllers, duplicate disks, duplicate network connections, etc. In a case, like you are describing where you create iSCSI storage on a single-node system, you are providing shared storage, but you have not provided for high availability. Your single-node iSCSI Target is a single point of failure. Loss of that single node would cause the whole cluster to come down.

Answer 4

What is your RPO/RTO requirements for DR?

Answer 5

Hello @Michael adams

Thank you for posting this concern on this community space.

I have read your case scenario description and I wonder if you have thought of using or considering the following feature...

https://learn.microsoft.com/en-us/azure-stack/hci/concepts/storage-spaces-direct-overview

Looking forward to your feedback,

Cheers,

Please "Accept the answer" if the information helped you. This will help us and others in the community as well.