For the longest time, when deploying a cluster with Windows Server, you only had the two options,
- Using a dedicated disk for the quorum, or
- Configuring an SMB file-share as the quorum witness
With Server 2016 and 2019, there is now a third option, Cloud Witness. The Cloud Witness leverages Azure Blob storage to provide that additional cluster/quorum vote.
Before showing you how this is done, one should understand the purpose of a witness/quorum is with respect to a failover cluster.
When one or more members of a cluster stops reporting to the other cluster members, there is a vote. The vote ensures that there is no split-vote, and ensures the cluster has a true owner. For example, in a two node cluster, if each node believe it is the owner, then this will cause a “split-brain”. In short, neither node will ever agree it is the owner (or not). This is where a quorum is required to determine who is the owner by providing the third vote, ie. majority. This ensures the cluster has a true owner by having the majority of votes. Each member gets a vote, plus the quorum.
Why this matters, in the even there is no quorum, a node from the cluster can be evicted and as a result will suspend all application services to prevent data corruption by more than one system writing data without the cluster services coordinating data writes and access. Depending on policies, VMs running on the ejected cluster member will either suspend operations or be migrated to other nodes before being ejected.
Below is a step-by-step guide on how to configure the Azure Blob storage as the Cloud Witness.
Assumptions:
- The Azure Blob storage account has already been created,
- The cluster with at least 2 nodes already exists.
Launch the PowerShell console as Administrator, and execute the following cmdlet:
Set-ClusterQuorum-CloudWitness-AccountName"storage_account_name"-AccessKey"primary_access_key"
Now if we go back to the Failover Manager console we can see we have successfully configured
In conclusion, deploying a Cloud Witness for a Failover Cluster is very simple, and in case of power outage in one datacenter, maintenance on a node, etc. then the entire cluster and its members (nodes) are all given an equal opportunity. Not only is it recommended and a requirement for 2-node clusters, but for any number of nodes, having a quorum is key ensuring high-availability. As mentioned, there are the traditional options such as using a dedicated disk or a file-share (SMB) as the cluster witness. However with Azure Blob storage with its 16×9 uptime, we can always ensure the quorum witness is online and available.