I'll make an attempt to keep it simple, and tell what is true in 'most'
cases. For some values of 'most' ;-)
The words used are at times confusing. "Block" mostly refers to
a logical filesystem block, which can be variable in size.
There's also "checksum" and "parity", which are completely
independent.
* The green and blue "blocks" shown in the diagram on page 11, do
the represent actual physical blocks on individual disks or a
single RAID-Z stripe write across multiple disks???
See Page 17: These are logical blocks, and can be variable in size.
* The parity for RAID-Z, where is it?? Surely not the checksum
stored together in the upper level direct and indirect block
pointer? And if not and it is written as a separate block on
another disks, then .......I am afraid I do not understand....
z-raid Parity vs zfs checksum
The parity is just a chunk of xor-ed data written for redundancy, and
is part of the same I/O transaction.
The checksum is a much smaller digest of the data used for detecting
the various modes of data corruption. This is what goes into the
metadata (logical) blocks above. A zfs file system always has checksums
and can function without parity.
* Could someone please elaborate more on the statement "Every block
is it's own RAID-Z stripe"??? The block being referred to is a
single block across multiple disks or a single disk?
If the storage pool will use an n-way raid-z configuration, the
(logical) block is first split into n-1 chunks, and an nth chunk
is added before any actual I/O takes place. Each chunk goes to
a separate disk.
This goes hand in hand with the answer to question 2. Because it's
Copy-on-Write, we only worry about new data when computing parity.
*My sincere apologies if the above questions seem trivial* . But I am
really struggling to reconcile the statement and the diagram.
Example
Logical block: (1 6k block of fs data) Could be any size <= 128k
|0|1|_|_|_|5|_|_|_|_|0|_| (12 x 512b blocks) --> ::checksum::
This is split into a single 4x 2k stripe:
3 chunks of 2k:
|00|01|02|03| --> disk1 (4 sectors)
|04|05|06|07| --> disk2 (4 sectors)
|08|09|10|11| --> disk3 (4 sectors)
1 chunk of parity:
|12|13|14|15| --> disk4 (4 sectors)
::checksum:: is then recorded in the metadata, which gets written
in a separate stripe. This is recursed for the metadata checksum,
until we reach the ueberblock, for which I won't explain the
redundancy and replication here.
Cheers,
Henk
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
