I'm working on migrating a customer's SQL Server 2019 cluster that's hosted on IaaS with Windows Server 2019. They chose the Standard E32-8ds v5 VM SKU in West Europe, but I'm hitting some roadblocks due to availability zone limits—the SKU is only available in one of the three zones in our subscription. I'm in touch with MS support and the capacity team to check if more capacity can be made available. Meanwhile, I've deployed two Standard E32-8ds v6 virtual machines with Premium SSD ZRS shared disks, but I'm facing issues in the Failover Cluster Manager because I can't use these shared disks. The Failover Cluster validation shows that the NVME disk controller isn't suitable for clustering, specifically saying, 'Disk bus type does not support clustering. Disk is on the system bus.' The customer is keen to stick with SQL Server 2019 and Windows Server 2019 due to tight deadlines. Does anyone have a workaround for this situation?
2 Answers
If I’m understanding correctly, you’re using Failover Cluster Instances instead of Availability Groups, right? My first question is—why not just use AGs? That would avoid needing to add the storage to the cluster at all. But if sticking with Failover Clustering is necessary, it seems there might not be a workaround for NVME issues with shared disks. You might want to check out the Microsoft docs if you haven't yet.
I've dealt with NVME disks and SQL before, though not the same exact problem. My temp disk was uninitialized, so I had to run a script to initialize and partition it. Eventually, I switched back to v4 and everything worked smoothly.
Yes, it's definitely Failover Cluster Instances. The customer prefers a like-for-like migration without changing the infrastructure.