[Freedombox-discuss] Switch to btrfs on the Freedombox?

[leen]

Hi Peter,

On 11/03/2014 12:29 PM, Petter Reinholdtsen wrote:
> 
> This weekend I finally was able to find a way for vmdebootstrap to
> create a bootable virtual Debian disk with root as a btrfs subvolume.
> The patch is available in <URL: https://bugs.debian.org/741223 >.
> 

Let me first say I think it's awesome what you are doing.

People working with btrfs is very useful. I've been following btrfs
mailinglist on and off to see if I can use it for some kind of
production use.

I'm sorry I didn't reply to your email earlier. I saw your email when I
was at work last week but I didn't have time to reply and forgot to send
in an email later on.

> The reason I have been working on this is because I believe a FreedomBox
> system would work better with a file and disk volume system that can be
> extended without remount and reboot, handle snapshots and provide
> resilience against block corruption.  As far as I know, the only option
> with these features included in Debian today is btrfs.  If zfs show up
> on Linux in Debian, it would be an option too.
> 
> This would allow us to plug new disks into the freedombox and add them
> to the storage system, and migrate data from the internal disk to the
> external disk (and back) if we want to.
> 
> This would also ensure blocks are checksummed all the way through the
> storage stack, making it possible to detect data corruption (and
> automatically correct for such corruption if a RAID like setup is
> configured) that would not be detected using ext3/4.
> 
> The btrfs file system is generally seen as sligthly unstable, and
> earlier could cause data loss.  I have not heard any such reports the
> last two years, but it should be part of the evaluation.  It earlier had
> problems when running out of disk space, but I have not heard about such
> problems lately.
> 

But let me point out that these errors can still occur:

"And btrfs can automatically allocate data and metadata chunks on demand
-- the catch is that it can't automatically unallocate
chunks on demand[1], a balance is required for that"

[1] Yet.  There's patches circulating that once thru discussion and
merged, should let btrfs automatically handle at least the normal cases
of data/metadata chunk imbalance.

http://thread.gmane.org/gmane.comp.file-systems.btrfs/39800/focus=39943

> So the question to discuss is, should we (when the patch in the bug
> report above) switch the freedom-maker scripts to create systems with
> btrfs?  Anyone against it?
> 

I'm not a freedombox developer, just an observer.