softraid —
software RAID
softraid0 at root
The
softraid device emulates a Host Bus Adapter
(HBA) that provides RAID and other I/O related services. The
softraid device provides a scaffold to implement
more complex I/O transformation disciplines. For example, one can tie chunks
together into a mirroring discipline. There really is no limit on what type of
discipline one can write as long as it fits the SCSI model.
softraid supports a number of
disciplines. A discipline is a collection of
functions that provides specific I/O functionality. This includes I/O path,
bring-up, failure recovery, and statistical information gathering. Essentially
a discipline is a lower level driver that provides the I/O transformation for
the softraid device.
A
volume is a virtual disk device that is made up
of a collection of chunks.
A
chunk is a partition or storage area of fstype
“RAID”.
disklabel(8) is used to alter
the fstype.
Currently
softraid supports the following
disciplines:
- RAID 0
- A striping discipline. It
segments data over a number of chunks to increase performance. RAID 0 does
not provide for data loss (redundancy).
- RAID 1
- A mirroring discipline. It
copies data across more than one chunk to provide for data loss. Read
performance is increased, though at the cost of write speed. Unlike
traditional RAID 1, softraid supports the use
of more than two chunks in a RAID 1 setup.
- RAID 5
- A striping discipline with
floating parity across all chunks. It stripes
data across chunks and provides parity to prevent data loss of a single
chunk failure. Read performance is increased; write performance does incur
additional overhead.
- CRYPTO
- An encrypting discipline. It
encrypts data on a single chunk to provide for data confidentiality.
CRYPTO does not provide redundancy.
- CONCAT
- A concatenating discipline.
It writes data to each chunk in sequence to provide increased capacity.
CONCAT does not provide redundancy.
installboot(8) may be used to
install
boot(8) in the boot
storage area of the
softraid volume. Boot support
is currently limited to the CRYPTO and RAID 1 disciplines on amd64, i386, and
sparc64 platforms. On sparc64, bootable chunks must be RAID partitions using
the letter ‘a’. At the
boot(8) prompt, softraid volumes
have names beginning with ‘sr’ and can be booted from like a
normal disk device. CRYPTO volumes will require a decryption passphrase or
keydisk at boot time.
An example to create a 3 chunk RAID 1 from scratch is as follows:
Initialize the partition tables of all disks:
# fdisk -iy wd1
# fdisk -iy wd2
# fdisk -iy wd3
Now create RAID partitions on all disks:
# printf "a\n\n\n\nRAID\nw\nq\n" | disklabel -E wd1
# printf "a\n\n\n\nRAID\nw\nq\n" | disklabel -E wd2
# printf "a\n\n\n\nRAID\nw\nq\n" | disklabel -E wd3
Assemble the RAID volume:
# bioctl -c 1 -l /dev/wd1a,/dev/wd2a,/dev/wd3a softraid0
The console will show what device was added to the system:
scsibus0 at softraid0: 1 targets
sd0 at scsibus0 targ 0 lun 0: <OPENBSD, SR RAID 1, 001> SCSI2
sd0: 1MB, 0 cyl, 255 head, 63 sec, 512 bytes/sec, 3714 sec total
It is good practice to wipe the front of the disk before using it:
# dd if=/dev/zero of=/dev/rsd0c bs=1m count=1
Initialize the partition table and create a filesystem on the new RAID volume:
# fdisk -iy sd0
# printf "a\n\n\n\n4.2BSD\nw\nq\n" | disklabel -E sd0
# newfs /dev/rsd0a
The RAID volume is now ready to be used as a normal disk device. See
bioctl(8) for more information
on configuration of RAID sets.
Install
boot(8) on the RAID volume:
At the
boot(8) prompt, load the /bsd
kernel from the RAID volume:
bio(4),
bioctl(8),
boot_sparc64(8),
disklabel(8),
fdisk(8),
installboot(8),
newfs(8)
The
softraid driver first appeared in
OpenBSD 4.2.
Marco Peereboom.
The driver relies on underlying hardware to properly fail chunks.
The RAID 1 discipline does not initialize the mirror upon creation. This is by
design because all sectors that are read are written first. There is no point
in wasting a lot of time syncing random data.
The RAID 5 discipline does not initialize parity upon creation, instead parity
is only updated upon write.
Currently there is no automated mechanism to recover from failed disks.
Certain RAID levels can protect against some data loss due to component failure.
RAID is
not a substitute for good backup
practices.
![[OpenBSD]](/openbsd.gif){kind=small}