On Fri, Feb 22, 2013 at 11:29:42AM -0500, Jerry Feldman wrote: > On 02/22/2013 11:01 AM, Rich Pieri wrote: > > On Fri, 22 Feb 2013 10:04:24 -0500 > > Jerry Feldman <g...@blu.org> wrote: > > > >> Most of the examples I have seen are to install btrfs on raw drives. > > Btrfs is, like ZFS, both file system and volume manager. There is > > typically no benefit to not allowing Btrfs to manage entire devices > > unless you need to have part of the disk not be Btrfs. Not allowing > > Btrfs to manage whole devices makes it more difficult to replace > > faulted devices. > > > >> redundant,, but your data is essentially stripped (RAID0) so you > >> effectively get more storage with the safety of RAID1. (You can > >> configure btrfs to be fully redundant if you want to). > > No. RAID0 means a device failure equals data loss. Mirrored metadata > > will not save you from that. What mirrored metadata gets you is a > > measure of protection against bit errors damaging file names, > > permissions, checksums and related information. > > > > Note that Btrfs mirrors data and metadata, not disks or disk blocks. A > > three-disk Btrfs raid1 is not three copies of every file extent. It is > > two copies of every file extent stored one each on two out of the three > > disks. > So, assume I have 2 physical volumes, /dev/sda and /dev/sdb. > mkfs.btrfs -d raid1 /dev/sda /dev/sdb > What happens if I get a failure on /dev/sdb. > Assume no snapshots.
Snapshots are irrelevant. -m raid1 is default, and you have added -d raid1, so everything is mirrored. When /dev/sdb drops out, you get an error message and everything continues working. Replace the disk and continue. -dsr- _______________________________________________ Discuss mailing list Discuss@blu.org http://lists.blu.org/mailman/listinfo/discuss
