I have been running two Azure Windows 2016 VM's, running SQL Server 2017, in a Failover Cluster setup.
For 3 weeks these have been running fine. But the last two days the Data Disk (connected as 'F:' where the SQL Data is stored) gets inexplicably unattached.
Thus the SQL Servers stop working, the failover is rendered pointless as it fails over to the other server which is also not working.
What would cause a Data Disk to detach like this?
Rebooting the VM's fixing the issue, thankfully. And all the data remains intact, but this can happen out of office hours and is really not ideal.
From your explanation the Disk wasn't detach, it sound like is was taken offline on the OS level.
Unfortunately this is a common thing for SQL Cluster to do when the disk is exhausted performance wise.
Make sure you are following Microsoft's Performance guidelines for SQL Server in Azure Virtual Machines
If you are following performance guidelines and still getting the issue then you need a support request to investigate performance on platform side.