What is the effect on LVM snapshot size when a file block is rewritten with it's original contents?
- by NevilleDNZ
I'm exploring using LVM snapshot's to off site incremental archives from a snapshot "master" file system. In essence: simply copy across only the files on the "master" that have changed since the last incremental copy to the "archive". Then snapshot the "archive" to retain the incremental.
I am a bit puzzled as to the block usage behaviour of the archive's own incremental snapshot.
I'm expecting that LVM is not smart enough to know that the "file block" is actually unchanged, and the a new copy will be allocated and written for the fresh "archive" file system.
Can anyone confirm this, or point me to a document/page that gives some hints?
BTW: the OS hard disk cache, hard disk physical cache and hard disk itself also doesn't need to do any actual "disk writes" as the "disk block" likewise is unnecessary. Any pointers to discussion of this style of optimisation would also be ineresting.