Re: lvm2/dm-thin snapshot memory and disk overhead

[Zdenek Kabelac <zdenek.kabelac@gmail.com>]

Dne 12. 10. 24 v 0:46 Drew Eckhardt napsal(a):
> What's the lvm2/dm-thin snapshot memory and disk overhead for each
> chunk (default 64k) overwritten?
> 
> 

Hi

Depends on the usage.

If you write full 64K block - the whole new provisioned chunk is written.

If you write 1st. 4K block to 'not yet provisioned' chunk for snapshot - then 
remains of 64-4 of the origin chunk to be 'copied' to this newly provisioned 
chunk.

If you write to already provisioned chunk for a snapshot thin LV - nothing is 
copied and all goes just the this chunks.

When origin 'write' to a chunk shared with 'snapshot' - then snapshot will 
keep the unmodified chunk and origin gets a new copy with updated content.

Note - snapshot thin LV is just like any other 'thin LV' - it just 'shares' 
already allocated chunks with its origin.

Simply said - writing task always 'pays' the price with block update...

Hopefully this makes it clear.

Regards

Zdenek


PS: it's way more efficient compared with old 'thick' snapshots...