I think the universal consensus is that outside of a very specific use case: multiple VDI desktops that share the same image, ZFS dedupe is completely useless at best and will destroy your dataset at worst by causing to be unmountable on any system that has less RAM than needed. In every other use case, the savings are not worth the trouble.
Even in the VDI use case, unless you have MANY copies of said disk images(like 5+ copies of each), it’s still not worth the increase in system resources needed to use ZFS dedupe.
It’s one of those “oooh shiny” nice features that everyone wants to use, but will regret it nearly every time.
I think the universal consensus is that outside of a very specific use case: multiple VDI desktops that share the same image, ZFS dedupe is completely useless at best and will destroy your dataset at worst by causing to be unmountable on any system that has less RAM than needed. In every other use case, the savings are not worth the trouble.
Even in the VDI use case, unless you have MANY copies of said disk images(like 5+ copies of each), it’s still not worth the increase in system resources needed to use ZFS dedupe.
It’s one of those “oooh shiny” nice features that everyone wants to use, but will regret it nearly every time.