So, I'm still at the playing around stage of ZFS at this point. I originally planned to use a ZVOL formatted to HFS+ for my non-admin user home directory (leaving my admin user on a native HFS+ boot volume SSD). I was able to get that to work pretty well, in that OS X El Capitan could point to the proper place for that non-admin user, my files were there when I logged in, all the software ran, etc. However... I ran into a showstopper for that approach. It appears that one can't make snapshots of a ZVOL while it is mounted and live; when I think about it, it makes perfect sense, since ZFS has no visibility into what HFS+ is doing with/to the volume. and it can't know a safe time to take the snapshot. I'm guessing that snapshots work just fine with a native ZFS dataset, since ZFS can choose to do the snapshot at an appropriate moment in between filesystem activity. Have I got that right?
So-o-o, I'm considering discarding the volume approach and just using a native ZFS filesystem dataset for my non-admin user home directory. I've followed the recommendations for creating an HFS+ compatible dataset from the wiki (having to do with case insensitivity, normalization, etc), and I've set the HFS mimic property to ON (I set all that up as pool properties so that they would be inherited by all child datasets).
My question is, how viable is this approach for the whole user directory under El Capitan? Does it step on the toes of SIP? Are there certain applications that don't really play nice unless a real HFS+ filesystem is used? Any listing of pitfalls for this use case would be greatly appreciated. Thanks,
Kurt