So, I'm in the process of replacing my spinning disks with an SSD, and I have a question about whether it is possible to either to transfer responsibility for a send from disk A to disk C, or else reverse the direction of a send.
Here is the scenario: I have a zpool 'tank' that is the master now, and I've been sending snapshots of dataset 'tank/ZHOME' to a pool on a backup disk 'ELITE'. They're just monthly snapshots (originally quarterly), so there are about a dozen of them that now exist on both the tank/ZHOME dataset, and the ELITE/ZHOME dataset. So far, so good.
Now I want to replace the spinning disks that host the 'tank' zpool with an SSD, but I would like to be able to continue to send snapshots from the new SSD to the same backup dataset 'ELITE/ZHOME'. Is it as simple as:
- make one final snapshot @April2018 on the tank/ZHOME
- zfs send @April2018 incrementally to 'ELITE/ZHOME'
- zfs send @April2018 (full send, not incremental) to SSD/ZHOME
- swap hardware around, switch my user directory to SSD/ZHOME
- do some work for a few weeks
- now take snapshot on SSD/ZHOME@May2018
- zfs send -i @April2018 SSD/ZHOME@May2018
That last line is intended to send incremental between the April snapshot and the new May snapshot, even though April was originally sourced from a different pool; in other words, does a receiving pool keep track of the unique ID of the originating pool, and refuse to accept any snapshot updates from any other pool? Or does a snapshot have a unique ID that travels along with it across sends, and allow it to accept an incremental snapshot send from a different source than the original, so long as the snapshot is preserved?
Another scenario that tests similar concepts: currently I'm doing sends from tank/ZHOME to ELITE/ZHOME; if I do a send from ELITE/ZHOME to SSD/ZHOME can I later turn around the direction, and do incremental sends back from SSD/ZHOME to ELITE/ZHOME as my ongoing backup strategy?
Thanks for your thoughts.