From btrfs Wiki
Jump to: navigation, search

On a simple level, use of a seed-backed device seems like a simple (possibly very much better integrated) unionfs/aufs-like filesystem layering. --Killermist 15:44, 19 May 2012 (UTC)

Starter logic:

When a device is made a "seed" device, and it is mounted again, it mounts read only.

btrfstune -S 1 /dev/parent
mount /dev/parent /mnt/test

Then the "child" device (for lack of a better concept) is added.

btrfs device add /dev/child /mnt/temp

/mnt/temp remains read-only until remounted read-write, at which time all writes go to /dev/child This, I understand.

Some questions:

1) Can /dev/parent be reused with another "child" device, such that /dev/parent is a seed for more than 1 "child"? --Killermist 15:44, 19 May 2012 (UTC)

2) Can /dev/parent be mounted read-write at some point later to update things requiring updating? And, if so, would those changes translate to the child devices that inherited /dev/parent as their seed, or would the "child devices" continue to inherit the instance of /dev/parent as it existed at time of creation of the "child"? --Killermist 15:44, 19 May 2012 (UTC)

3) Can seeding be nested? For example there is an original seed, a "grandparent" so to speak. Add (for a time) a writable "parent" (child of grandparent). Convert "parent" to a seed. Then add a "child" (child of parent). Possibly expand to grand-child, great-grand-child, etc relationships. (though such use-cases seem unlikely, but you never know) --Killermist 15:44, 19 May 2012 (UTC)

Some answers: --Ebonacco

1) yes, personally tested with

kernel 3.8.8-1.el6.elrepo.x86_64 #1 SMP and btrfs-progs-0.20-0.2.git91d9eec.el6.x86_64

Personal tools