On 2010-03-15 16:50, Khyron:
Yeah, this threw me. A 3 disk RAID-Z2 doesn't make sense, because at a
redundancy level, RAID-Z2 looks like RAID 6. That is, there are 2 levels of
parity for the data. Out of 3 disks, the equivalent of 2 disks will be used
to store redundancy (parity) data and only 1 disk equivalent will store actual
data. This is what others might term a "degenerate case of 3-way
mirroring", except with a lot more computational overhead since we're
performing 2
parity calculations.
I'm curious what the purpose of creating a 3 disk RAID-Z2 pool is/was?
(For my own personal edification. Maybe there is something for me to learn
from this example.)
One reason for a raid-z2 would be to protect yourself against a
structural firmware error corrupting certain data patterns in a similar
way. An alternate pattern could make it to and from disk correctly.
I once visited a customer who was experiencing data corruption caused by
their firmware (back in 1991 I guess). They had no reproducable case
and neither of us was in anyway satisfied. Only half a year later did I
hear that this particular type of disk needed a firmware update, which
came to light in a DBMS - which would force a read of its data after
every write.
Apparently it took so long because the erroneous behaviour only happened
under certain conditions.
Cheers,
Henk
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
