Btrfs and COW

[thenonsense]

Hi all,

 

I'm trying to understand something about Btrfs and how it tracks its multiple copies of files.  I have a couple scenarios and I'm tryting to understand how COW resolves with them:

1. A file is put onto the file system by User A.  User B updates the file.  What differentiates User A from User B?  How is data chosen (original vs updated) to present to User A or B?  What about User C, who has seen neither original nor updated variant?
2. A library, or video game, or even a git repo is put onto the file system.  That library / game / repo receives an update 2 weeks later.  Large amounts of edits may be made, so does this rapidly consume storage?
3. A file is updated/deleted by User A.  How does that file get updated/deleted in the original, so that User B either sees the new version or no file at all?

I'd appreciate if someone could explain to me some of these items or point me in a direction that helps to explain.  I've been over wikipedia a couple times, and devouring stackoverflow, but I'm still not seeing clear answers to my questions, likely out of not understanding the answers.

[JorgeB]

Not sure what you are asking exactly, for that purpose a COW filesystem works as any other, once a file is modified any user will only see the new version of that file, only if you have snaphosts (or reflinks) you can access previous versions.

[thenonsense]

I figured it out.  I had expected CoW to be a lot more intensive than it actually is.  It's not.  Copies of files are implicitly made only when a file is copied and then one of those copies is modified upon.  I expected some copy-in-place transactional craziness that versioned files as they were modified, and didn't care if a copy was made previously.  I'm confident I'm not explaining my original point of view well, but that's because it was pretty stupid to begin with.