Adding a new Cache device and you setup your cluster with -CacheDeviceModel and now the new device wont be used as Cache?

Hello Everyone, it’s been a while since my last blog post. So i thought id share something i came over in our Azure Stack HCI(Former S2D) slack channel. https://azurestackhci.slack.com

Someone said that they had a new Node that when they added that node to the cluster, the new Cache devices would not bind to the Capacity drives on that cluster. The reason for this is when the cluster was created, it was run with Enable-ClusterS2D -CacheDeviceModel. When doing so you tell the cluster that the only allowed Cache device is this model.

This set’s the drive model in an XML file that is in the Storage Spaces Direct health service. So to allow other cache devices this needs to be updated. On a cluster where the -CacheDeviceModel is not set, the xml file is empty allowing any device of same “type” to be used as cache.

The recommended solution for an already set CacheDeviceModel is to update the XML file with the new Cache device. Microsoft has a blog post about it here. You can also specify what capacity drives that are supported to run on this cluster.

Let me show you an example of what you can do. Run this command to get the output.

Get-StorageSubSystem Cluster* | Get-StorageHealthSetting -Name "System.Storage.SupportedComponents.Document"

This will get you this output, this is an example from a running system, your cache devices will be a diffrent model.

[cce lang=”xml”] MZ7KM960HMJP0D3 [/cce]

To add another cache device, you simply put in the new one.

[cce lang=”xml”] MZ7KM960HMJP0D3 SSDSC2KG960G7R [/cce]

Once you have done this you can import the new XML file to your health service by running

$MyXML = Get-Content Filepath | Out-String  
Get-StorageSubSystem Cluster* | Set-StorageHealthSetting -Name "System.Storage.SupportedComponents.Document" -Value $MyXML 

Now you should be able to set the Cache device as Journal

If you wan’t to add just specefic Capacity drives use the Disks setting instead.