When talking about ZFS, we often talk about problems that some people dismiss as remote ... like triple-failure in a RAID6 or data-loss due to hardware bugs. But those problems happen really - for example as observed by Storagebod. The more important your data gets, the more possible sh.ts you have to get out of the equation. Beginning from "RAID is sufficient" if you can recreate the data, but you don't have the time to get the data from a backup up to "multi-site replication, multi-site backup with different products, daily restore tests and putting your newly created tapes into a safe location after testing the tapes".

A few years ago i would have said there is a class "No protection needed" but given that printed photos came out of fashion, this stage doesn't exist anymore. Or do you want to explain your significant other that you just lost all your photos since your first vacation? I would rather delete the databases of a SAP system.

And by the way: There is a basic law in those projects when it's the first larger backup/restore project. The customer wants the full monty as described at first. After creating the matching quote, the customer mostly discloses, that he just wants to pay for "RAID is sufficient". At the end, the customer gets what he really needs, albeit he has to pay more than he initially wanted. Fortunately at the second time, all parties save time by more resonable assumptions But that is the reason why my colleagues doing such projects are called "consultants" instead of "quotants"

Read More about [Sh.t happens - really ......