For the duplicates ZAP , since the pool survives it if you just delete it, isn't it possible to reconstruct it? For the unique zap you can run through all blocks and those that don't exist were probably in here but got pruned. So you can re add them. But by that logic it will also find duplicates eventually. So this should be recoverable. Also keeping a pruned list let's you add an offline dedup, which goes at disk speed looking for duplicates instead of ram speed. And so can dedup already written blocks later. And a command to say "hey, dedup this and this file's blocks because I know they're nearly identical" could be used to speed that process up.
For the duplicates ZAP , since the pool survives it if you just delete it, isn't it possible to reconstruct it? For the unique zap you can run through all blocks and those that don't exist were probably in here but got pruned. So you can re add them. But by that logic it will also find duplicates eventually.
So this should be recoverable.
Also keeping a pruned list let's you add an offline dedup, which goes at disk speed looking for duplicates instead of ram speed. And so can dedup already written blocks later.
And a command to say "hey, dedup this and this file's blocks because I know they're nearly identical" could be used to speed that process up.
Thanks Allan, great presentation.
Will we see one on zRAID expansion too?
The RAIDZ Expansion talk has been posted: th-cam.com/video/tqyNHyq0LYM/w-d-xo.html