Ias - Question about ZFS snapshots

New About Yours API Help
1.8 KB, Plain text
Hi Allen and Benedict,

I have a question about ZFS.  At the current time I only do snapshots of 
my datasets manually.  Then I manually zfs send/receive them to a backup 
pool.  I'd like to keep all the snapshots going as far back as possible 
on the backup pool, since it has plenty of space.  However, I'd like to 
thin out some of the snapshots on the source pool for space reasons. 
I'm just wondering if there are any things I need to look out for if I 
do this...

Suppose my dataset is named MyData and I have snapshots @January, 
@February, @March, and @April on both the source and backup pool.

I could theoretically delete @January, @February, and @March snapshots 
on the source, generate a new @May snapshot, and then send an 
incremental update from @April through @May to the backup, right?  You 
could summarize this question as "do I only need the latest two 
snapshots to be able to send an incremental update via zfs send/receive?"

Here's a trickier scenario: suppose I started with the same four 
original snapshots, but then destroyed @February and @March on the 
source dataset to thin out the source pool a bit, leaving @January and 
@April.  Then suppose a lightning strike took out both of the disks in 
my backup pool, but left my source pool undamaged (unlikely, I know, but 
perhaps it was turned off at the time).  Now I want to create a new 
backup pool on new hardware, and send it all the data that still exists 
on the source.  Of course I can send the entire @January snapshot to the 
new backup, no problem.  Can I still send an incremental backup of 
@January through @April successfully?  You could summarize this question 
as "will an incremental send/receive still work if you destroy some 
intermediate snapshots between them?"

Thank you for your thoughts and caveats,

Kurt
Pasted 2 months, 2 weeks ago — Expires in 285 days
URL: http://dpaste.com/1CC86MX