I understand his decision to skip the HW RAID controller (I like mdadm too). But a BBU is definitely critical to this op. NVMe has so many more queues that on a busy box it's going to hold way too much uncommitted data and metadata. Now the only other reasonable option is a UPS but it adds significant RUs/cost to the setup.
I am no expert, but are you sure that uncommitted data on the ssd is an issue?
The Intel DC P3600 has a built in capacitor that should give it econoff backup power to commit the data. The SanDisk Extreme Pro disks don't have volatile cache, but instead uses SLC NAND flash for cashing, that will survive a power loss.
There is of course still the issue that data send to the server and stored in main memory will be lost if you loses power, but a HW RAID controller with battery backup would not have prevented that (but an UPS of course might).
It would definitely prevent an uncommitted data problem if fsync or O_DIRECT were used (which should always be used for critical writes).
UPSes defend against power outages, but they're only one rung on the data-integrity ladder. Controller BBUs, on the other hand, protect against both power outages AND kernel panics.
I tend to agree with you here runarb, I truly don't believe there is much that a hardware RAID contrôlée and battery will protect you from with modern flash storage, at this point I think time is better spent on improving / tuning the database and filesystem for resiliency.
I went down the same road - using mdadm and a UPS instead of a HW RAID controller with battery backup unit. LSI MegaRAID SAS 9261-8i SGL with a battery backup costs roughly 650 EUR; Eaton 9130i 1500VA Rack 2U UPS costs 850 EUR. So it's a bit more expensive, but you would be buying that UPS anyway.
