if using BTRFileSystem

Started by ralul, 2012/07/03, 14:57:17

Previous topic - Next topic

agaida

System Backup:


cp -a source backup


System Restore:


cp -a backup source


:twisted:
There's this special biologist word we use for "stable". It's "dead". ~ Jack Cohen

bad_aptitude

Is the time almost upon us that siduction could use something like "apt-btrfs-snapshot" (unfortunately its only a Ubuntu hack at the moment) to manage the rolling back of gnarly dist-upgrades. As valuable as the siduction utlility for managing kernels is, it only solves half of the potential problems associated with dist upgrades.
I'm not a programmer - but I would be happy to test such a feature.

Regards
bad-aptitude

agaida

if you found one who develop and test this and it work reliable we would be happy to distribute such a package.

EDIT: Reliability is the point. Such a tool should not base on a experimental filesystem.
There's this special biologist word we use for "stable". It's "dead". ~ Jack Cohen

michaa7

Quote from: "agaida"System Backup:


cp -a source backup


I'd suggest a slightly enhanced version of this command:

cp -ax /path/to/<source_partition>/. /path/to/<backup>/


The "-x" option prevents "cp" to dive into linked partitions, and the "/." at the end of the source path makes sure only the content is copied (not the containing folder).
Ok, you can't code, but you still might be able to write a bug report for Debian's sake

agaida

good point, i'm a litte bit lazy some times
There's this special biologist word we use for "stable". It's "dead". ~ Jack Cohen

bad_aptitude

One advantage, using the snapshot feature of BTRFS would have over
cp -a source backup
is the snapshot of a BTRFS directory is effectively a set of hardlinks to the original files. So when taking a snapshot prior to a dist-upgrade on BTRFS, only the files changed by the dist-upgrade are "duplicated".  At least that is my understanding of the process.
It would seem that this approach could save a lot of time and space.

agaida

when btrfs will be stable - in 5 or 10 years - i would eventually agree
There's this special biologist word we use for "stable". It's "dead". ~ Jack Cohen

bad_aptitude

But isn't stable just "dead"?

agaida

let me cite from the unix-hater-handbook:
"Sure - the new filesystem will corrupt your data. But look, how fast it is!"

The Clue with btrfs is: I tested it a while ago, it was fucking slow at writes and i decided that i don't like it on my main machines. Imho btrfs is a wonderful thing with netbooks and small notebook - machines, where you mostly need high reading speed. Given that it is reliable and have reliable tools for crash recovery.
There's this special biologist word we use for "stable". It's "dead". ~ Jack Cohen