xref: /netbsd-src/share/man/man4/raid.4 (revision 21e37cc72a480a47828990a439cde7ac9ffaf0c6)
1.\"     $NetBSD: raid.4,v 1.29 2004/03/05 14:43:01 wiz 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 February 29, 2004
64.Dt RAID 4
65.Os
66.Sh NAME
67.Nm raid
68.Nd RAIDframe disk driver
69.Sh SYNOPSIS
70.Cd options RAID_AUTOCONFIG
71.Cd options RAID_DIAGNOSTIC
72.Cd options RF_ACC_TRACE=n
73.Cd options RF_DEBUG_MAP=n
74.Cd options RF_DEBUG_PSS=n
75.Cd options RF_DEBUG_QUEUE=n
76.Cd options RF_DEBUG_QUIESCE=n
77.Cd options RF_DEBUG_RECON=n
78.Cd options RF_DEBUG_STRIPELOCK=n
79.Cd options RF_DEBUG_VALIDATE_DAG=n
80.Cd options RF_DEBUG_VERIFYPARITY=n
81.Cd options RF_INCLUDE_CHAINDECLUSTER=n
82.Cd options RF_INCLUDE_EVENODD=n
83.Cd options RF_INCLUDE_INTERDECLUSTER=n
84.Cd options RF_INCLUDE_PARITY_DECLUSTERING=n
85.Cd options RF_INCLUDE_PARITY_DECLUSTERING_DS=n
86.Cd options RF_INCLUDE_PARITYLOGGING=n
87.Cd options RF_INCLUDE_RAID5_RS=n
88.Pp
89.Cd "pseudo-device raid" Op Ar count
90.Sh DESCRIPTION
91The
92.Nm
93driver provides RAID 0, 1, 4, and 5 (and more!) capabilities to
94.Nx .
95This
96document assumes that the reader has at least some familiarity with RAID
97and RAID concepts.  The reader is also assumed to know how to configure
98disks and pseudo-devices into kernels, how to generate kernels, and how
99to partition disks.
100.Pp
101RAIDframe provides a number of different RAID levels including:
102.Bl -tag -width indent
103.It RAID 0
104provides simple data striping across the components.
105.It RAID 1
106provides mirroring.
107.It RAID 4
108provides data striping across the components, with parity
109stored on a dedicated drive (in this case, the last component).
110.It RAID 5
111provides data striping across the components, with parity
112distributed across all the components.
113.El
114.Pp
115There are a wide variety of other RAID levels supported by RAIDframe.
116The configuration file options to enable them are briefly outlined
117at the end of this section.
118.Pp
119Depending on the parity level configured, the device driver can
120support the failure of component drives.  The number of failures
121allowed depends on the parity level selected.  If the driver is able
122to handle drive failures, and a drive does fail, then the system is
123operating in "degraded mode".  In this mode, all missing data must be
124reconstructed from the data and parity present on the other
125components.  This results in much slower data accesses, but
126does mean that a failure need not bring the system to a complete halt.
127.Pp
128The RAID driver supports and enforces the use of
129.Sq component labels .
130A
131.Sq component label
132contains important information about the component, including a
133user-specified serial number, the row and column of that component in
134the RAID set, and whether the data (and parity) on the component is
135.Sq clean .
136If the driver determines that the labels are very inconsistent with
137respect to each other (e.g. two or more serial numbers do not match)
138or that the component label is not consistent with its assigned place
139in the set (e.g. the component label claims the component should be
140the 3rd one in a 6-disk set, but the RAID set has it as the 3rd component
141in a 5-disk set) then the device will fail to configure.  If the
142driver determines that exactly one component label seems to be
143incorrect, and the RAID set is being configured as a set that supports
144a single failure, then the RAID set will be allowed to configure, but
145the incorrectly labeled component will be marked as
146.Sq failed ,
147and the RAID set will begin operation in degraded mode.
148If all of the components are consistent among themselves, the RAID set
149will configure normally.
150.Pp
151Component labels are also used to support the auto-detection and
152autoconfiguration of RAID sets.  A RAID set can be flagged as
153autoconfigurable, in which case it will be configured automatically
154during the kernel boot process.  RAID file systems which are
155automatically configured are also eligible to be the root file system.
156There is currently only limited support (alpha and pmax architectures)
157for booting a kernel directly from a RAID 1 set, and no support for
158booting from any other RAID sets.  To use a RAID set as the root
159file system, a kernel is usually obtained from a small non-RAID
160partition, after which any autoconfiguring RAID set can be used for the
161root file system.  See
162.Xr raidctl 8
163for more information on autoconfiguration of RAID sets.
164Note that with autoconfiguration of RAID sets, it is no longer
165necessary to hard-code SCSI IDs of drives.
166The autoconfiguration code will
167correctly configure a device even after any number of the components
168have had their device IDs changed or device names changed.
169.Pp
170The driver supports
171.Sq hot spares ,
172disks which are on-line, but are not
173actively used in an existing file system.  Should a disk fail, the
174driver is capable of reconstructing the failed disk onto a hot spare
175or back onto a replacement drive.
176If the components are hot swappable, the failed disk can then be
177removed, a new disk put in its place, and a copyback operation
178performed.  The copyback operation, as its name indicates, will copy
179the reconstructed data from the hot spare to the previously failed
180(and now replaced) disk.  Hot spares can also be hot-added using
181.Xr raidctl 8 .
182.Pp
183If a component cannot be detected when the RAID device is configured,
184that component will be simply marked as 'failed'.
185.Pp
186The user-land utility for doing all
187.Nm
188configuration and other operations
189is
190.Xr raidctl 8 .
191Most importantly,
192.Xr raidctl 8
193must be used with the
194.Fl i
195option to initialize all RAID sets.  In particular, this
196initialization includes re-building the parity data.  This rebuilding
197of parity data is also required when either a) a new RAID device is
198brought up for the first time or b) after an un-clean shutdown of a
199RAID device.  By using the
200.Fl P
201option to
202.Xr raidctl 8 ,
203and performing this on-demand recomputation of all parity
204before doing a
205.Xr fsck 8
206or a
207.Xr newfs 8 ,
208file system integrity and parity integrity can be ensured.  It bears
209repeating again that parity recomputation is
210.Ar required
211before any file systems are created or used on the RAID device.  If the
212parity is not correct, then missing data cannot be correctly recovered.
213.Pp
214RAID levels may be combined in a hierarchical fashion.  For example, a RAID 0
215device can be constructed out of a number of RAID 5 devices (which, in turn,
216may be constructed out of the physical disks, or of other RAID devices).
217.Pp
218The first step to using the
219.Nm
220driver is to ensure that it is suitably configured in the kernel.  This is
221done by adding a line similar to:
222.Bd -unfilled -offset indent
223pseudo-device   raid   4       # RAIDframe disk device
224.Ed
225.Pp
226to the kernel configuration file.  The
227.Sq count
228argument (
229.Sq 4 ,
230in this case), specifies the number of RAIDframe drivers to configure.
231To turn on component auto-detection and autoconfiguration of RAID
232sets, simply add:
233.Bd -unfilled -offset indent
234options RAID_AUTOCONFIG
235.Ed
236.Pp
237to the kernel configuration file.
238.Pp
239All component partitions must be of the type
240.Dv FS_BSDFFS
241(e.g. 4.2BSD) or
242.Dv FS_RAID .
243The use of the latter is strongly encouraged, and is required if
244autoconfiguration of the RAID set is desired.  Since RAIDframe leaves
245room for disklabels, RAID components can be simply raw disks, or
246partitions which use an entire disk.
247.Pp
248A more detailed treatment of actually using a
249.Nm
250device is found in
251.Xr raidctl 8 .
252It is highly recommended that the steps to reconstruct, copyback, and
253re-compute parity are well understood by the system administrator(s)
254.Ar before
255a component failure.  Doing the wrong thing when a component fails may
256result in data loss.
257.Pp
258Additional internal consistency checking can be enabled by specifying:
259.Bd -unfilled -offset indent
260options RAID_DIAGNOSTIC
261.Ed
262.Pp
263These assertions are disabled by default in order to improve
264performance.
265.Pp
266RAIDframe supports an access tracing facility for tracking both
267requests made and performance of various parts of the RAID systems
268as the request is processed.
269To enable this tracing the following option may be specified:
270.Bd -unfilled -offset indent
271options RF_ACC_TRACE=1
272.Ed
273.Pp
274For extensive debugging there are a number of kernel options which
275will aid in performing extra diagnosis of various parts of the
276RAIDframe sub-systems.
277Note that in order to make full use of these options it is often
278necessary to enable one or more debugging options as listed in
279.Pa src/sys/dev/raidframe/rf_options.h .
280As well, these options are also only typically useful for people who wish
281to debug various parts of RAIDframe.
282The options include:
283.Pp
284For debugging the code which maps RAID addresses to physical
285addresses:
286.Bd -unfilled -offset indent
287options RF_DEBUG_MAP=1
288.Ed
289.Pp
290Parity stripe status debugging is enabled with:
291.Bd -unfilled -offset indent
292options RF_DEBUG_PSS=1
293.Ed
294.Pp
295Additional debugging for queuing is enabled with:
296.Bd -unfilled -offset indent
297options RF_DEBUG_QUEUE=1
298.Ed
299.Pp
300Problems with non-quiescent file systems should be easier to debug if
301the following is enabled:
302.Bd -unfilled -offset indent
303options RF_DEBUG_QUIESCE=1
304.Ed
305.Pp
306Stripelock debugging is enabled with:
307.Bd -unfilled -offset indent
308options RF_DEBUG_STRIPELOCK=1
309.Ed
310.Pp
311Additional diagnostic checks during reconstruction are enabled with:
312.Bd -unfilled -offset indent
313options RF_DEBUG_RECON=1
314.Ed
315.Pp
316Validation of the DAGs (Directed Acyclic Graphs) used to describe an
317I/O access can be performed when the following is enabled:
318.Bd -unfilled -offset indent
319options RF_DEBUG_VALIDATE_DAG=1
320.Ed
321.Pp
322Additional diagnostics during parity verification are enabled with:
323.Bd -unfilled -offset indent
324options RF_DEBUG_VERIFYPARITY=1
325.Ed
326.Pp
327There are a number of less commonly used RAID levels supported by
328RAIDframe.
329These additional RAID types should be considered experimental, and
330may not be ready for production use.
331The various types and the options to enable them are shown here:
332.Pp
333For Even-Odd parity:
334.Bd -unfilled -offset indent
335options RF_INCLUDE_EVENODD=1
336.Ed
337.Pp
338For RAID level 5 with rotated sparing:
339.Bd -unfilled -offset indent
340options RF_INCLUDE_RAID5_RS=1
341.Ed
342.Pp
343For Parity Logging (highly experimental):
344.Bd -unfilled -offset indent
345options RF_INCLUDE_PARITYLOGGING=1
346.Ed
347.Pp
348For Chain Declustering:
349.Bd -unfilled -offset indent
350options RF_INCLUDE_CHAINDECLUSTER=1
351.Ed
352.Pp
353For Interleaved Declustering:
354.Bd -unfilled -offset indent
355options RF_INCLUDE_INTERDECLUSTER=1
356.Ed
357.Pp
358For Parity Declustering:
359.Bd -unfilled -offset indent
360options RF_INCLUDE_PARITY_DECLUSTERING=1
361.Ed
362.Pp
363For Parity Declustering with Distributed Spares:
364.Bd -unfilled -offset indent
365options RF_INCLUDE_PARITY_DECLUSTERING_DS=1
366.Ed
367.Pp
368The reader is referred to the RAIDframe documentation mentioned in the
369.Sx HISTORY
370section for more detail on these various RAID configurations.
371.Sh WARNINGS
372Certain RAID levels (1, 4, 5, 6, and others) can protect against some
373data loss due to component failure.  However the loss of two
374components of a RAID 4 or 5 system, or the loss of a single component
375of a RAID 0 system, will result in the entire file systems on that RAID
376device being lost.
377RAID is
378.Ar NOT
379a substitute for good backup practices.
380.Pp
381Recomputation of parity
382.Ar MUST
383be performed whenever there is a chance that it may have been
384compromised.  This includes after system crashes, or before a RAID
385device has been used for the first time.  Failure to keep parity
386correct will be catastrophic should a component ever fail -- it is
387better to use RAID 0 and get the additional space and speed, than it
388is to use parity, but not keep the parity correct.  At least with RAID
3890 there is no perception of increased data security.
390.Sh FILES
391.Bl -tag -width /dev/XXrXraidX -compact
392.It Pa /dev/{,r}raid*
393.Nm
394device special files.
395.El
396.Sh SEE ALSO
397.Xr sd 4 ,
398.Xr MAKEDEV 8 ,
399.Xr config 8 ,
400.Xr fsck 8 ,
401.Xr mount 8 ,
402.Xr newfs 8 ,
403.Xr raidctl 8
404.Sh HISTORY
405The
406.Nm
407driver in
408.Nx
409is a port of RAIDframe, a framework for rapid prototyping of RAID
410structures developed by the folks at the Parallel Data Laboratory at
411Carnegie Mellon University (CMU).  RAIDframe, as originally distributed
412by CMU, provides a RAID simulator for a number of different
413architectures, and a user-level device driver and a kernel device
414driver for Digital Unix.  The
415.Nm
416driver is a kernelized version of RAIDframe v1.1.
417.Pp
418A more complete description of the internals and functionality of
419RAIDframe is found in the paper "RAIDframe: A Rapid Prototyping Tool
420for RAID Systems", by William V. Courtright II, Garth Gibson, Mark
421Holland, LeAnn Neal Reilly, and Jim Zelenka, and published by the
422Parallel Data Laboratory of Carnegie Mellon University.
423The
424.Nm
425driver first appeared in
426.Nx 1.4 .
427.Sh COPYRIGHT
428.Bd -unfilled
429The RAIDframe Copyright is as follows:
430.Pp
431Copyright (c) 1994-1996 Carnegie-Mellon University.
432All rights reserved.
433.Pp
434Permission to use, copy, modify and distribute this software and
435its documentation is hereby granted, provided that both the copyright
436notice and this permission notice appear in all copies of the
437software, derivative works or modified versions, and any portions
438thereof, and that both notices appear in supporting documentation.
439.Pp
440CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
441CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
442FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
443.Pp
444Carnegie Mellon requests users of this software to return to
445.Pp
446 Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
447 School of Computer Science
448 Carnegie Mellon University
449 Pittsburgh PA 15213-3890
450.Pp
451any improvements or extensions that they make and grant Carnegie the
452rights to redistribute these changes.
453.Ed
454