Not to start a war, I swear.
How is a breaking change, such as this (which was committed in a “minor” update), considered suitable for enterprise clients?
“Swap on a Linux base has been giving us issues! Let’s just gut everything to do with swap, which means we’ll no longer partition a couple GB for drives being added to a new vdev. What could possibly go wrong?”
It comes off as a lack of QA.
As humble home user myself, even I could foresee this problem.
In a chapter about ZFS, FreeBSD’s handbook recommends to “provision” (partition) a drive, to allow for some buffer, for the explicit reason of not finding yourself stuck with this problem when the time comes to replace disks in a pool. (This has nothing to do with swap. Such a “buffer space” of a few GB should be a default standard.)
iXsystems can’t simply rely on their trust of HDD/SSD manufacturers. Even when a user ordered the same size and same model of drive, the new ones were slightly smaller than the existing drives in the pool. Now he is out of luck, unless he returns the drive and then purchases a larger (and more expensive) capacity drive.