Hello. Suppose you: _ use UEFI; _ boot from ZFS; _ are on 12.3; _ checkpoint the root pool; _ upgrade to 13.1. Now 13.1's loader.efi won't be able to find your pool and the machine won't boot. If you are lucky you'll be able to see "can not read checkpoint data", but this message might scroll fast out of sight. It would not be a big problem, if this was documented. N.B. The above checklist reflect my case, but it could possibly be relaxed (e.g. WRT system versions) or narrowed (pool version or features?).
I need more specific details on how to recreate this. Why can't loader.efi see it? What are the characteristics of the ZFS snapshot that gives the loader grief? Does this persist in a zfs send -> zfs receieve? If so can I get that dataset? How important is this to you working?
(In reply to Warner Losh from comment #1) First off, there are no "characteristics of the ZFS snapshot": we are talking about "zpool checkpoint", not "zfs snap". I have no idea on why loader.efi can't boot from a checkpointed ZFS pool: I only saw the message above ("can not read checkpoint data"). I don't have the original problematic pool anymore: it was a production server and I needed to boot it ASAP, so I just issued "zpool checkpoint -d" (that is, as soon as I relized what the problem was). I guess, in order to recreate the situation, procedure would be: install 12.3 UEFI+ZFS, checkpoint, upgrade to 13.1 (maybe upgrading boot loader is enough). The only thing I could add is it was a zraid5 pool (3 disks); not sure it matters. Importance is "less than bulk" to me: it would have saved me three hours of spreading panic if I had known; now I'll simply check if a checkpoint exists before upgrading (or, if I forget, boot from an USB key and remove it later). My only goal was to let other people know, so maybe they won't be hit so hard by this.