sraid(7) (RM400 only) sraid(7)
NAME
sraid - SCSI RAID interface on an RM400
SYNOPSIS
/dev/iosb/[r]sraidbcullsp
DESCRIPTION
sraid provides SCSI magnetic RAID devices. IOS SCSI RAID devices are
accessed through the special files in the directory /dev/ios0. The
naming convention for all IOS RAID devices is as follows (all numbers
are specified in decimal):
/dev/iosb/[r]sraidbcullsp
b Bus number: always 0.
r Designates the raw (character) device.
c The decimal controller number of the SCSI host adaptor. The con-
troller number is either 0 or 1 on bus 0 or 0 to 3 on bus 1.
u The device's unit number (address). The address may range from 0
to 6.
l The logical unit number (LUN). This number may range from 0 to 6.
p Designates the LUN slice (partition) (0 to 15).
MAJOR-MINOR DEVICE NUMBER
The major-minor device number can be determined via /sbin/autoconf -d
devicename. The meaning of the number returned is described below.
The major-minor number is a 32-bit number formatted as follows:
BBBMMMMM MMMMMMaa aaaccbdd duuupppp
B Bus number: 0 or 1.
M The major device number.
a The host adaptor number on the specified bus, always 0.
c The host adaptor's controller number.
b The SCSI bus the device is attached to on the specified host
adaptor. Since each host adaptor has only one SCSI bus, b is
always 0.
d The SCSI device address (0 to 6) on the specified SCSI bus.
Page 1 Reliant UNIX 5.44 Printed 11/98
sraid(7) (RM400 only) sraid(7)
u The SCSI Logical Unit Number of the specified SCSI device, if
supported. Most devices do not use LUN's, therefore bits 4-6 will
usually be 0.
p The partition number (0 to 15).
ERROR DESCRIPTION
The error messages displayed for the IOS RAID drives have the follow-
ing formats:
ios0/sraid010l4s1: Write Extended 32 <seek> [3506:0:0] <dkblk> 721024 <pblk> 4824256
<status> 0b950200 (hard 1/1) Aborted Command - A drive channel was reset
(RAID) Extended Sense Bytes 13-101 for lpb at 0xc04d5c00
* 00020000 00300002 02450000 00000402 00040000 00031b11 00001202 2a800049
* 9cc00000 20000000 0f315433 35303130 31323100 00000000 00303130 30010400
* 00354434 3143525a 32382020 20202028 43292044 45430000 05000000
CMN: WARNING: cmn 38477 ios0/sraid010l4 Field Replaceable Unit (FRU) belongs to:
DRIVE GROUP
Failure report for drive at controller SCSI channel 3, SCSI ID 5:
List of failed drive components:
Time 14:59:16 up 1:55
CMN: WARNING: cmn 52643 ios0/sraid010l4 state changed to Degraded RAID5 Lun
ios0/sraid010l4s1: Write Extended 32 <seek> [3506:0:0] <dkblk> 721024 <pblk> 4824256
<status> 063f0200 (hard 2/2) Unit Attention - Caution: drive replace required!
(RAID) Extended Sense Bytes 13-101 for lpb at 0xc04d5c00
* 80350000 00000001 00000000 00000000 00000000 00000000 00000400 2a800049
* 9cc00000 20000000 0f315433 35303130 31323100 00000000 00303130 30020401
* 35124434 3143525a 32382020 20202028 43292044 45430000 05000000
The second error message will be described below:
ios0/sraid010l4s1:
The IOS RAID drive in error. In this case, the RAID LUN located
logical controller 1, drive 0, lun 4, in partition 1.
<cmd>
The SCSI command in error. In this case, a write of 32 physical
sectors was attempted.
<seek>
The seek address for the write described by
[cylinder:head:sector]. Note that the seek address is determined
in one of two ways. If the SCSI request sense command extended
sense information bytes are valid, then this value is decoded
into cylinder, head, and sector values. Otherwise, the start
address specified in (this case by) the SCSI write command is
decoded. In either case, the seek address does not take into
account any remapping that might have been done previously by
format. So remapping is not based upon the seek address.
Page 2 Reliant UNIX 5.44 Printed 11/98
sraid(7) (RM400 only) sraid(7)
<dkblk>
The 512-byte LUN block start number (offset from the current par-
tition). In this case the eight sector write started at block 16
in partition 8.
<pblk>
The physical LUN block number. Note that the physical LUN block
number is determined in one of two ways. If the SCSI request
sense command extended sense information bytes are valid, then
this value is used. Otherwise, the start address specified in
(this case by) the SCSI write command is used. Use this value for
specifying a start address in format.
<status> KKAASSCC
The IOS composite status code. The composite status code has the
following format:
KK The SCSI Sense Key value.
AA The SCSI additional Sense Key value.
SS The SCSI status returned to the SCSI host adaptor.
CC The IOS host adaptor error code.
(hard 1/1)
The error type, followed by the retry count. There are two error
types: hard and soft. Soft errors are retried. Unrecoverable soft
errors and errors that should not be retried are classified as
hard. The retry count is expressed as the number of retries
divided by the maximum retries allowed. In the case of hard
errors, the value is always 1/1.
Unit Not Ready
The decoded SCSI Sense Key value message.
LUN not ready, START UNIT required
The decoded SCSI Additional Sense Key value message.
(RAID)
The device type (RAID).
Extended Sense Bytes
The byte address of the saved extended sense bytes displayed in
the next lines after this message. In this case, bytes 12 to 135
of the MS12 LUN's extended sense data were saved after sending
the SCSI request sense command after the error on the write SCSI
command.
lpb at
The kernel address of the local parameter block for the SCSI
write command issued to the RAID LUN.
Page 3 Reliant UNIX 5.44 Printed 11/98
sraid(7) (RM400 only) sraid(7)
* 80350000 ....
The extended sense data. The amount of data displayed depends
upon the number of saved extended sense bytes specified for the
device type's profile in /etc/conf/pack.d/sri/space.c.
SCSI RAID CONFIGURATION
SCSI devices and RAID LUNs are identified and initialized during boot
time in the SCSI drive probe. A RAID LUN is not actually configured,
however, until it has been opened. During system boot, a RAID device
is identified solely on the basis of its SCSI inquiry name. Once a LUN
has been opened, and SCSI geometry information has been retrieved, the
device is configured. RAID LUN configuration may be checked at any
time via the autoconf command. Below is sample autoconf -l output:
ios0/sraid070l0 MS12(4087MB) -- Optimal RAID5 LUN
ios0/sraid070l1 MS12: -- not configured
ios0/sraid070l2 MS12(3067MB) -- partition not set
Following the LUN device name are two fields: the type and the current
LUN state. So, ios0/sraid070l0 has a type name of MS12, and a LUN
state of optimal RAID5. LUNs which haven't been configured yet display
a shortened form of the SCSI device type name followed by an unconfig-
ured message. RAID LUNs which are partially configured (i.e. dkindex
hasn't been run yet), will show partition not set. See dkindex(8) and
dkpsdfinit(8) for more information on configuring LUNs.
After drive spin up, the LUN format, geometry and logical unit parame-
ters are retrieved through the mode sense SCSI command. This step
occurs when the drive is first opened. The LUN format and geometry
parameters are used to determine the physical characteristics of the
LUN (for example, cylinders, heads, sectors, skew, and so forth). The
logical unit parameters are used to determine the RAID level, the
current state of the LUN and the number of physical LUN which are
displayed by autoconf or used internally in the kernel. The MS12 sup-
ports the RAID Levels: 0,1,3 and 5. This information is used for error
handling in the driver and displayed by autoconf. The LUN state can
have one of the following: Optimal, Degraded, Reconstructing or Dead.
A degraded or dead LUN requires manually intervention, which can be
done with apr(1M) and replacing a failed physical drive.
The reserved cylinder count is further adjusted by one cylinder for a
reserved cylinder in the data area. This area contains the disk index
structure [see psdf(4)], and is read at configuration time. The com-
mand dksetup needs to be run to initialize a RAID LUN for use.
Finally, one more reserved cylinder for a CE cylinder in the data area
is added. Since the native CE cylinder requires many gyrations to
access, or is inaccessible to the mortal user, a CE cylinder in the
data area is reserved for the Format Information Table produced by
format. This contains information as to when the RAID LUN was last
formatted, drive serial number, and basic geometry information. The
defect list is maintained by the SCSI RAID controller, and is not
Page 4 Reliant UNIX 5.44 Printed 11/98
sraid(7) (RM400 only) sraid(7)
stored physically on the RAID LUN.
SCSI RAID NOTES
The physical sector size is not necessarily 512 bytes and is dependent
on the RAID level and the number of physical disks. The SCSI RAID
driver converts the block count into the number of sectors required
for a 512 byte transfer. When errors occur both the physical block and
the system block size are printed. The formatter expects that the phy-
sical block will be provided, should a user wish to map out a bad
block.
The LUNs of the RAID system can be configured and examined by acf(1M).
Use the autoconf command to verify any configuration changes.
SEE ALSO
acf(1M), apr(1M), psdf(4), autoconf(8), dkindex(8), dkpart(8),
dkpsdfinit(8), dksetup(8), format(8), setinfo(8).
Page 5 Reliant UNIX 5.44 Printed 11/98