I am working with a company right now that is running Thin VMs on thin backend storage. They would like to maintain no more than 20% overallocation on any datastore, but are encountering inconsistent sDRS migrations they attributed to sDRS seeing only the used space rather than the allocated space.
I recall that DRS includes a buffer of 25% for both memory and storage overallocation, using the PercentIdleMBinSpaceDemand and PercentIdleMBinMemDemand options. I've read this article, accompanied with the Storage DRS FAQ KB (which leads me to believe the article is still considered valid), which indicate that if I set the PercentIdleMBinSpaceDemand to 100, sDRS will consider the entirety of a VMs allocated space in considering recommendations. However, when I set this on a datastore cluster that has several datastores way over-provisioned (15TB provisioned on 6TB actual space), with several more almost empty, sDRS still does not come back with any recommendations. Am I missing anything, or am I misunderstanding how this should work?