ZFS Syncoid : merge two sets of snapshots together (feature request)

[MynaITLabs]

Would be handy to help merge two sets of ZFS snapshots together into one clean separate filesystem with all snapshots.

For example,
If there are 3 servers, each have the same ZFS version.
MAIN, BACKUP1, BACKUP2

• Syncoid does a daily and hourly snapshot MAIN server data files /main-pool/data (in ZFS filesystem main-pool/data)
• Syncoid is daily transferring snapshots main-pool to BACKUP1 zfs pool backup1-pool/main-pool/data. (i.e. offsite backup)
• Snapshots older than 30 days are removed. (Since they are actually on the BACKUP1 server)
  After 3 months, offsite server room fire damages the computer and SSD drives. Bummer !
• Replacement and data recovery will take at least 42 days.
  So I create a new server BACKUP2.
• Syncoid is now daily transferring snapshots MAIN 'main-pool' to BACKUP2 zfs pool backup2-pool/main-pool/data.
  After 69 days, BACKUP1 SSD drive data has been recovered, and back up and running
  I now have two older snapshot sets that I want to merge together into one complete set.
• BACKUP1 backup1-pool/mail-pool/data (first 90 days of MAIN)
• BACKUP2 backup2-pool/mail-pool/data (day 91-160 of MAIN)

One method is make a list of all snapshots and sort by date, then create a new filesystem based on the oldest to newest.

Would be great if syncoid could take several parameters and do the 'hand-waving with smoke-and-mirrors' automagically.
Something like...
$ syncoid --merge-source BACKUP1:backup1-pool/mail-pool/data --merge-source BACKUP2:backup2-pool/mail-pool/data --merge-dest BACKUP1:backup1-pool/mail-pool/data

NOTE:

• if --merge-dest matches any of the sources, then would need a temp working area. Then when complete process rename the DEST to DEST_PREV and DEST_TEMP to DEST.

[MynaITLabs]

For a more complex examples:

• BACKUP1 and BACKUP2 actually had a mix of different days on them and not exactly sequential to each other ?
• There are several different sources of similar snapshots (BACKUP1, BACKUP2, MAIN, BACKUP3, ZAPHOD, etc) merged into one dataset filesystem ?

[jimsalterjrs]

What you're asking for isn't really possible in ZFS, let alone sanoid/syncoid.

The closest thing to what you're looking for would be to make certain you do NOT delete the last common snapshot MAIN1 had with BACKUP1 before you send BACKUP1 off to Drivesavers or whoever. Then you build your BACKUP2. Then when BACKUP1 returns from the drive recovery outfit--which, honestly, it's probably not going to in the way that you think it is; they typically are only going to recover files from the main filesystem, not restore every snapshot--but assuming you did, in fact, get your actual original BACKUP1 system back with its pool intact, you could then simply replicate BACKUP2--which will still contain the last common snapshot between MAIN1 and BACKUP1--onto BACKUP1.

At that point, you just back up directly from MAIN1 to BACKUP1 again, and you tear down BACKUP2 for parts, repurpose it for something else, whatever.

Your second idea, of mixing and matching snapshots that are "not exactly sequential" and "on a mix of different days"--is literally impossible to do with ZFS replication, sorry.

[memnochproxy]

thanks for your input on this, Jim