1.\" $NetBSD: raid.4,v 1.15 2000/05/13 15:22:18 mycroft Exp $ 2.\" 3.\" Copyright (c) 1998 The NetBSD Foundation, Inc. 4.\" All rights reserved. 5.\" 6.\" This code is derived from software contributed to The NetBSD Foundation 7.\" by Greg Oster 8.\" 9.\" Redistribution and use in source and binary forms, with or without 10.\" modification, are permitted provided that the following conditions 11.\" are met: 12.\" 1. Redistributions of source code must retain the above copyright 13.\" notice, this list of conditions and the following disclaimer. 14.\" 2. Redistributions in binary form must reproduce the above copyright 15.\" notice, this list of conditions and the following disclaimer in the 16.\" documentation and/or other materials provided with the distribution. 17.\" 3. All advertising materials mentioning features or use of this software 18.\" must display the following acknowledgement: 19.\" This product includes software developed by the NetBSD 20.\" Foundation, Inc. and its contributors. 21.\" 4. Neither the name of The NetBSD Foundation nor the names of its 22.\" contributors may be used to endorse or promote products derived 23.\" from this software without specific prior written permission. 24.\" 25.\" THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 26.\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 27.\" TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 28.\" PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 29.\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 30.\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 31.\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 32.\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 33.\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 34.\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 35.\" POSSIBILITY OF SUCH DAMAGE. 36.\" 37.\" 38.\" Copyright (c) 1995 Carnegie-Mellon University. 39.\" All rights reserved. 40.\" 41.\" Author: Mark Holland 42.\" 43.\" Permission to use, copy, modify and distribute this software and 44.\" its documentation is hereby granted, provided that both the copyright 45.\" notice and this permission notice appear in all copies of the 46.\" software, derivative works or modified versions, and any portions 47.\" thereof, and that both notices appear in supporting documentation. 48.\" 49.\" CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" 50.\" CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND 51.\" FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. 52.\" 53.\" Carnegie Mellon requests users of this software to return to 54.\" 55.\" Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU 56.\" School of Computer Science 57.\" Carnegie Mellon University 58.\" Pittsburgh PA 15213-3890 59.\" 60.\" any improvements or extensions that they make and grant Carnegie the 61.\" rights to redistribute these changes. 62.\" 63.Dd November 9, 1998 64.Dt RAID 4 65.Os 66.Sh NAME 67.Nm raid 68.Nd RAIDframe disk driver 69.Sh SYNOPSIS 70.Cd "pseudo-device raid" Op Ar count 71.Sh DESCRIPTION 72The 73.Nm 74driver provides RAID 0, 1, 4, and 5 (and more!) capabilities to NetBSD. This 75document assumes that the reader has at least some familiarity with RAID 76and RAID concepts. The reader is also assumed to know how to configure 77disks and pseudo-devices into kernels, how to generate kernels, and how 78to partition disks. 79.Pp 80RAIDframe provides a number of different RAID levels including: 81.Bl -tag -width indent 82.It RAID 0 83provides simple data striping across the components. 84.It RAID 1 85provides mirroring. 86.It RAID 4 87provides data striping across the components, with parity 88stored on a dedicated drive (in this case, the last component). 89.It RAID 5 90provides data striping across the components, with parity 91distributed across all the components. 92.El 93.Pp 94There are a wide variety of other RAID levels supported by RAIDframe, 95including Even-Odd parity, RAID level 5 with rotated sparing, Chained 96declustering, and Interleaved declustering. The reader is referred 97to the RAIDframe documentation mentioned in the 98.Sx HISTORY 99section for more detail on these various RAID configurations. 100.Pp 101Depending on the parity level configured, the device driver can 102support the failure of component drives. The number of failures 103allowed depends on the parity level selected. If the driver is able 104to handle drive failures, and a drive does fail, then the system is 105operating in "degraded mode". In this mode, all missing data must be 106reconstructed from the data and parity present on the other 107components. This results in much slower data accesses, but 108does mean that a failure need not bring the system to a complete halt. 109.Pp 110The RAID driver supports and enforces the use of 111.Sq component labels . 112A 113.Sq component label 114contains important information about the component, including a 115user-specified serial number, the row and column of that component in 116the RAID set, and whether the data (and parity) on the component is 117.Sq clean . 118If the driver determines that the labels are very inconsistent with 119respect to each other (e.g. two or more serial numbers do not match) 120or that the component label is not consistent with it's assigned place 121in the set (e.g. the component label claims the component should be 122the 3rd one a 6-disk set, but the RAID set has it as the 3rd component 123in a 5-disk set) then the device will fail to configure. If the 124driver determines that exactly one component label seems to be 125incorrect, and the RAID set is being configured as a set that supports 126a single failure, then the RAID set will be allowed to configure, but 127the incorrectly labeled component will be marked as 128.Sq failed , 129and the RAID set will begin operation in degraded mode. 130If all of the components are consistent among themselves, the RAID set 131will configure normally. 132.Pp 133Component labels are also used to support the auto-detection and 134auto-configuration of RAID sets. A RAID set can be flagged as 135auto-configurable, in which case it will be configured automatically 136during the kernel boot process. RAID filesystems which are 137automatically configured are also eligible to be the root filesystem. 138While there is no support for booting directly from a RAID set, it is 139possible to boot from a small partition which contains a kernel, and 140have the root filesystem on a RAID set. See 141.Xr raidctl 8 142for more information on auto-configuration of RAID sets. 143.Pp 144The driver supports 145.Sq hot spares , 146disks which are on-line, but are not 147actively used in an existing filesystem. Should a disk fail, the 148driver is capable of reconstructing the failed disk onto a hot spare 149or back onto a replacement drive. 150If the components are hot swapable, the failed disk can then be 151removed, a new disk put in its place, and a copyback operation 152performed. The copyback operation, as its name indicates, will copy 153the reconstructed data from the hot spare to the previously failed 154(and now replaced) disk. Hot spares can also be hot-added using 155.Xr raidctl 8 . 156.Pp 157If a component cannot be detected when the RAID device is configured, 158that component will be simply marked as 'failed'. 159.Pp 160The user-land utility for doing all 161.Nm 162configuration and other operations 163is 164.Xr raidctl 8 . 165Most importantly, 166.Xr raidctl 8 167must be used with the 168.Fl i 169option to initialize all RAID sets. In particular, this 170initialization includes re-building the parity data. This rebuilding 171of parity data is also required when either a) a new RAID device is 172brought up for the first time or b) after an un-clean shutdown of a 173RAID device. By using the 174.Fl P 175option to 176.Xr raidctl 8 , 177and performing this on-demand recomputation of all parity 178before doing a 179.Xr fsck 8 180or a 181.Xr newfs 8 , 182filesystem integrity and parity integrity can be ensured. It bears 183repeating again that parity recomputation is 184.Ar required 185before any filesystems are created or used on the RAID device. If the 186parity is not correct, then missing data cannot be correctly recovered. 187.Pp 188RAID levels may be combined in a hierarchical fashion. For example, a RAID 0 189device can be constructed out of a number of RAID 5 devices (which, in turn, 190may be constructed out of the physical disks, or of other RAID devices). 191.Pp 192It is important that drives be hard-coded at their respective 193addresses (i.e. not left free-floating, where a drive with SCSI ID of 1944 can end up as /dev/sd0c) for well-behaved functioning of the RAID 195device. This is true for all types of drives, including IDE, HP-IB, 196etc. For normal SCSI drives, for example, the following can be used 197to fix the device addresses: 198.Bd -unfilled -offset indent 199sd0 at scsibus0 target 0 lun ? # SCSI disk drives 200sd1 at scsibus0 target 1 lun ? # SCSI disk drives 201sd2 at scsibus0 target 2 lun ? # SCSI disk drives 202sd3 at scsibus0 target 3 lun ? # SCSI disk drives 203sd4 at scsibus0 target 4 lun ? # SCSI disk drives 204sd5 at scsibus0 target 5 lun ? # SCSI disk drives 205sd6 at scsibus0 target 6 lun ? # SCSI disk drives 206.Ed 207.Pp 208See 209.Xr sd 4 210for more information. The rationale for fixing the device addresses 211is as follows: Consider a system with three SCSI drives at SCSI ID's 2124, 5, and 6, and which map to components /dev/sd0e, /dev/sd1e, and 213/dev/sd2e of a RAID 5 set. If the drive with SCSI ID 5 fails, and the 214system reboots, the old /dev/sd2e will show up as /dev/sd1e. The RAID 215driver is able to detect that component positions have changed, and 216will not allow normal configuration. If the device addresses are hard 217coded, however, the RAID driver would detect that the middle component 218is unavailable, and bring the RAID 5 set up in degraded mode. Note 219that the auto-detection and auto-configuration code does not care 220about where the components live. The auto-configuration code will 221correctly configure a device even after any number of the components 222have been re-arranged. 223.Pp 224The first step to using the 225.Nm 226driver is to ensure that it is suitably configured in the kernel. This is 227done by adding a line similar to: 228.Bd -unfilled -offset indent 229pseudo-device raid 4 # RAIDframe disk device 230.Ed 231.Pp 232to the kernel configuration file. The 233.Sq count 234argument ( 235.Sq 4 , 236in this case), specifies the number of RAIDframe drivers to configure. 237To turn on component auto-detection and auto-configuration of RAID 238sets, simply add: 239.Bd -unfilled -offset indent 240options RAID_AUTOCONFIG 241.Ed 242.Pp 243to the kernel configuration file. 244.Pp 245All component partitions must be of the type 246.Dv FS_BSDFFS 247(e.g. 4.2BSD) or 248.Dv FS_RAID . 249The use of the latter is strongly encouraged, and is required if 250auto-configuration of the RAID set is desired. Since RAIDframe leaves 251room for disklabels, RAID components can be simply raw disks, or 252partitions which use an entire disk. 253.Pp 254A more detailed treatment of actually using a 255.Nm 256device is found in 257.Xr raidctl 8 . 258It is highly recommended that the steps to reconstruct, copyback, and 259re-compute parity are well understood by the system administrator(s) 260.Ar before 261a component failure. Doing the wrong thing when a component fails may 262result in data loss. 263.Pp 264.Sh WARNINGS 265Certain RAID levels (1, 4, 5, 6, and others) can protect against some 266data loss due to component failure. However the loss of two 267components of a RAID 4 or 5 system, or the loss of a single component 268of a RAID 0 system, will result in the entire filesystems on that RAID 269device being lost. 270RAID is 271.Ar NOT 272a substitute for good backup practices. 273.Pp 274Recomputation of parity 275.Ar MUST 276be performed whenever there is a chance that it may have been 277compromised. This includes after system crashes, or before a RAID 278device has been used for the first time. Failure to keep parity 279correct will be catastrophic should a component ever fail -- it is 280better to use RAID 0 and get the additional space and speed, than it 281is to use parity, but not keep the parity correct. At least with RAID 2820 there is no perception of increased data security. 283.Pp 284.Sh FILES 285.Bl -tag -width /dev/XXrXraidX -compact 286.It Pa /dev/{,r}raid* 287.Nm 288device special files. 289.El 290.Pp 291.Sh SEE ALSO 292.Xr MAKEDEV 8 , 293.Xr raidctl 8 , 294.Xr config 8 , 295.Xr fsck 8 , 296.Xr mount 8 , 297.Xr newfs 8 , 298.Xr sd 4 299.Sh HISTORY 300The 301.Nm 302driver in 303.Nx 304is a port of RAIDframe, a framework for rapid prototyping of RAID 305structures developed by the folks at the Parallel Data Laboratory at 306Carnegie Mellon University (CMU). RAIDframe, as originally distributed 307by CMU, provides a RAID simulator for a number of different 308architectures, and a user-level device driver and a kernel device 309driver for Digital Unix. The 310.Nm 311driver is a kernelized version of RAIDframe v1.1. 312.Pp 313A more complete description of the internals and functionality of 314RAIDframe is found in the paper "RAIDframe: A Rapid Prototyping Tool 315for RAID Systems", by William V. Courtright II, Garth Gibson, Mark 316Holland, LeAnn Neal Reilly, and Jim Zelenka, and published by the 317Parallel Data Laboratory of Carnegie Mellon University. 318The 319.Nm 320driver first appeared in 321.Nx 1.4 . 322.Sh COPYRIGHT 323.Bd -unfilled 324The RAIDframe Copyright is as follows: 325 326Copyright (c) 1994-1996 Carnegie-Mellon University. 327All rights reserved. 328 329Permission to use, copy, modify and distribute this software and 330its documentation is hereby granted, provided that both the copyright 331notice and this permission notice appear in all copies of the 332software, derivative works or modified versions, and any portions 333thereof, and that both notices appear in supporting documentation. 334 335CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" 336CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND 337FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. 338 339Carnegie Mellon requests users of this software to return to 340 341 Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU 342 School of Computer Science 343 Carnegie Mellon University 344 Pittsburgh PA 15213-3890 345 346any improvements or extensions that they make and grant Carnegie the 347rights to redistribute these changes. 348.Ed 349