We currently have a 48 node hybrid cluster running our Horizon VDI environment with non persistent desktops. The biggest issue is Oulook and OST files. I'm not allowed to turn off cached mode, so I'm redirecting users OST files to network shares. We currently have about 2,600 users and I have those OST files spread across 7 servers.

We brought in a consultant with more Nutanix experience than I have and it was suggested that we convert some nodes to all flash, and we would use Affinity Rules to pin these OST file servers to the AF nodes. Makes perfect sense to me based on Nutanix using localization for data.

So we have purchased the correct drives to convert 8 nodes to all flash. Storage capacity is going to be almost a wash, I think we are going to drop about 10 TB out of our 200TB logical.

I'm in the process of removing the first 4 nodes from the cluster so we can convert them and I got an error from Prism that's very concerning. I know this is a false positive today, but it's exactly what we planned on doing. The error is:





Cause: All-flash nodes and non-all-flash nodes are put into the same cluster

Impact: The cluster may experience performance degradation or instability.

Resolution: Remove either all-flash nodes or non-all-flash nodes from the cluster







Can anyone give some insight to this error and why this would be? I could understand if I had a bunch of AF nodes and added some hybrid that my performance would degrade, but I don't understand why having hybrid and adding AF would be a bad thing...