Has anyone successfully implemented any of the following Persistent Memory use scenarios with any version of SQL Server that is deployed as a Failover Cluster Instance (FCI) that is using shared SAN for data/log files in a Windows Failover Cluster? And for these PMEM devices, I'm referring to Intel OPTANE 256GB DIMMS, likely 4 per failover host. And in reality, I'm doing this in a VM, but Windows supports vPMEM at this point, so that isn't my concern here.
- Temp DB on DAX enabled PMEM Disk
- Temp DB on a BLOCK mode enabled PMEM Disk
- Persistent Log Buffer (aka Tail of the Log) feature using DAX enabled PMEM Disk
- Hybrid Buffer Pool feature using DAX enabled PMEM Disk
- Any other PMEM "Enlightenment" modes/ use-cases
I've see that #1 and #2 is technically supported, but a failed Temp DB dedicated storage device failure will not fail the cluster instance to the secondary node.
For #3, I was thinking we could implement a script to add the Persistent Log Buffer enabled log file to each database at startup of SQL Server, specifying a local path that is consistent across FCI nodes. But I feel this may not be "supported". I'm also not 100% confident if something level of log data would be lost if a power loss invoked failover occurred vs. a clean failover.
For #4, I don't see any means to make that feature work with a FCI, only in a AG Group or stand-alone install, which is too bad.
For #5, I'd love to hear what you may have experienced.
Thank you.
