[Discuss] On Btrfs raid and odd-count disks

[richard.pieri at gmail.com (Richard Pieri)]

To elaborate on this:

Given three disks: A, B and C, each 500G in a three-disk raid 1
configuration.

If you write a 100G file that file is written to disks A and B. None of
the file data is written to C. If you then write a 50G file then the
file data will be written to disks C and A. None of the file data is
written to disk B. At this point, disk A has 150G used, Disk B has 100G
used and disk C has 50G used. The largest file that you can write is
350G, limited by the available capacity on disk A.

In a three disk RAID 1e configuration, the 100G file will have file data
spread across all three disks. Likewise the 50G file. The largest file
that you can write is 600g: (1500G/2)-150G because data blocks are
striped across all three disks in the set.

RAID 1e is marketing gobbledygook. The spiel goes something like this:
if you have N disks and you lose up to (N/2)-1 (or something like that)
then you might not lose your data.

RAID is not about data integrity or preventing data loss. RAID is about
no single point of failure. That is, a single failure will not knock you
out. If you are concerned about two or more failures then you should be
looking at nested RAID such as RAID 100 or RAID 50.

-- 
Rich P.