1.\" $NetBSD: raidctl.8,v 1.61 2010/01/27 09:26:16 wiz Exp $ 2.\" 3.\" Copyright (c) 1998, 2002 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.\" 18.\" THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 19.\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 20.\" TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 21.\" PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 22.\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 23.\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 24.\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 25.\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 26.\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 27.\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 28.\" POSSIBILITY OF SUCH DAMAGE. 29.\" 30.\" 31.\" Copyright (c) 1995 Carnegie-Mellon University. 32.\" All rights reserved. 33.\" 34.\" Author: Mark Holland 35.\" 36.\" Permission to use, copy, modify and distribute this software and 37.\" its documentation is hereby granted, provided that both the copyright 38.\" notice and this permission notice appear in all copies of the 39.\" software, derivative works or modified versions, and any portions 40.\" thereof, and that both notices appear in supporting documentation. 41.\" 42.\" CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" 43.\" CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND 44.\" FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. 45.\" 46.\" Carnegie Mellon requests users of this software to return to 47.\" 48.\" Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU 49.\" School of Computer Science 50.\" Carnegie Mellon University 51.\" Pittsburgh PA 15213-3890 52.\" 53.\" any improvements or extensions that they make and grant Carnegie the 54.\" rights to redistribute these changes. 55.\" 56.Dd January 27, 2010 57.Dt RAIDCTL 8 58.Os 59.Sh NAME 60.Nm raidctl 61.Nd configuration utility for the RAIDframe disk driver 62.Sh SYNOPSIS 63.Nm 64.Op Fl v 65.Fl a Ar component Ar dev 66.Nm 67.Op Fl v 68.Fl A Op yes | no | root 69.Ar dev 70.Nm 71.Op Fl v 72.Fl B Ar dev 73.Nm 74.Op Fl v 75.Fl c Ar config_file Ar dev 76.Nm 77.Op Fl v 78.Fl C Ar config_file Ar dev 79.Nm 80.Op Fl v 81.Fl f Ar component Ar dev 82.Nm 83.Op Fl v 84.Fl F Ar component Ar dev 85.Nm 86.Op Fl v 87.Fl g Ar component Ar dev 88.Nm 89.Op Fl v 90.Fl G Ar dev 91.Nm 92.Op Fl v 93.Fl i Ar dev 94.Nm 95.Op Fl v 96.Fl I Ar serial_number Ar dev 97.Nm 98.Op Fl v 99.Fl m Ar dev 100.Nm 101.Op Fl v 102.Fl M 103.Oo yes | no | set 104.Ar params 105.Oc 106.Ar dev 107.Nm 108.Op Fl v 109.Fl p Ar dev 110.Nm 111.Op Fl v 112.Fl P Ar dev 113.Nm 114.Op Fl v 115.Fl r Ar component Ar dev 116.Nm 117.Op Fl v 118.Fl R Ar component Ar dev 119.Nm 120.Op Fl v 121.Fl s Ar dev 122.Nm 123.Op Fl v 124.Fl S Ar dev 125.Nm 126.Op Fl v 127.Fl u Ar dev 128.Sh DESCRIPTION 129.Nm 130is the user-land control program for 131.Xr raid 4 , 132the RAIDframe disk device. 133.Nm 134is primarily used to dynamically configure and unconfigure RAIDframe disk 135devices. 136For more information about the RAIDframe disk device, see 137.Xr raid 4 . 138.Pp 139This document assumes the reader has at least rudimentary knowledge of 140RAID and RAID concepts. 141.Pp 142The command-line options for 143.Nm 144are as follows: 145.Bl -tag -width indent 146.It Fl a Ar component Ar dev 147Add 148.Ar component 149as a hot spare for the device 150.Ar dev . 151Component labels (which identify the location of a given 152component within a particular RAID set) are automatically added to the 153hot spare after it has been used and are not required for 154.Ar component 155before it is used. 156.It Fl A Ic yes Ar dev 157Make the RAID set auto-configurable. 158The RAID set will be automatically configured at boot 159.Ar before 160the root file system is mounted. 161Note that all components of the set must be of type 162.Dv RAID 163in the disklabel. 164.It Fl A Ic no Ar dev 165Turn off auto-configuration for the RAID set. 166.It Fl A Ic root Ar dev 167Make the RAID set auto-configurable, and also mark the set as being 168eligible to be the root partition. 169A RAID set configured this way will 170.Ar override 171the use of the boot disk as the root device. 172All components of the set must be of type 173.Dv RAID 174in the disklabel. 175Note that only certain architectures 176.Pq currently alpha, i386, pmax, sparc, sparc64, and vax 177support booting a kernel directly from a RAID set. 178.It Fl B Ar dev 179Initiate a copyback of reconstructed data from a spare disk to 180its original disk. 181This is performed after a component has failed, 182and the failed drive has been reconstructed onto a spare drive. 183.It Fl c Ar config_file Ar dev 184Configure the RAIDframe device 185.Ar dev 186according to the configuration given in 187.Ar config_file . 188A description of the contents of 189.Ar config_file 190is given later. 191.It Fl C Ar config_file Ar dev 192As for 193.Fl c , 194but forces the configuration to take place. 195Fatal errors due to uninitialized components are ignored. 196This is required the first time a RAID set is configured. 197.It Fl f Ar component Ar dev 198This marks the specified 199.Ar component 200as having failed, but does not initiate a reconstruction of that component. 201.It Fl F Ar component Ar dev 202Fails the specified 203.Ar component 204of the device, and immediately begin a reconstruction of the failed 205disk onto an available hot spare. 206This is one of the mechanisms used to start 207the reconstruction process if a component does have a hardware failure. 208.It Fl g Ar component Ar dev 209Get the component label for the specified component. 210.It Fl G Ar dev 211Generate the configuration of the RAIDframe device in a format suitable for 212use with the 213.Fl c 214or 215.Fl C 216options. 217.It Fl i Ar dev 218Initialize the RAID device. 219In particular, (re-)write the parity on the selected device. 220This 221.Em MUST 222be done for 223.Em all 224RAID sets before the RAID device is labeled and before 225file systems are created on the RAID device. 226.It Fl I Ar serial_number Ar dev 227Initialize the component labels on each component of the device. 228.Ar serial_number 229is used as one of the keys in determining whether a 230particular set of components belong to the same RAID set. 231While not strictly enforced, different serial numbers should be used for 232different RAID sets. 233This step 234.Em MUST 235be performed when a new RAID set is created. 236.It Fl m Ar dev 237Display status information about the parity map on the RAID set, if any. 238If used with 239.Fl v 240then the current contents of the parity map will be output (in 241hexadecimal format) as well. 242.It Fl M Ic yes Ar dev 243.\"XXX should there be a section with more info on the parity map feature? 244Enable the use of a parity map on the RAID set; this is the default, 245and greatly reduces the time taken to check parity after unclean 246shutdowns at the cost of some very slight overhead during normal 247operation. 248Changes to this setting will take effect the next time the set is 249configured. 250Note that RAID-0 sets, having no parity, will not use a parity map in 251any case. 252.It Fl M Ic no Ar dev 253Disable the use of a parity map on the RAID set; doing this is not 254recommended. 255This will take effect the next time the set is configured. 256.It Fl M Ic set Ar cooldown Ar tickms Ar regions Ar dev 257Alter the parameters of the parity map; parameters to leave unchanged 258can be given as 0, and trailing zeroes may be omitted. 259.\"XXX should this explanation be deferred to another section as well? 260The RAID set is divided into 261.Ar regions 262regions; each region is marked dirty for at most 263.Ar cooldown 264intervals of 265.Ar tickms 266milliseconds each after a write to it, and at least 267.Ar cooldown 268\- 1 such intervals. 269Changes to 270.Ar regions 271take effect the next time is configured, while changes to the other 272parameters are applied immediately. 273The default parameters are expected to be reasonable for most workloads. 274.It Fl p Ar dev 275Check the status of the parity on the RAID set. 276Displays a status message, 277and returns successfully if the parity is up-to-date. 278.It Fl P Ar dev 279Check the status of the parity on the RAID set, and initialize 280(re-write) the parity if the parity is not known to be up-to-date. 281This is normally used after a system crash (and before a 282.Xr fsck 8 ) 283to ensure the integrity of the parity. 284.It Fl r Ar component Ar dev 285Remove the spare disk specified by 286.Ar component 287from the set of available spare components. 288.It Fl R Ar component Ar dev 289Fails the specified 290.Ar component , 291if necessary, and immediately begins a reconstruction back to 292.Ar component . 293This is useful for reconstructing back onto a component after 294it has been replaced following a failure. 295.It Fl s Ar dev 296Display the status of the RAIDframe device for each of the components 297and spares. 298.It Fl S Ar dev 299Check the status of parity re-writing, component reconstruction, and 300component copyback. 301The output indicates the amount of progress 302achieved in each of these areas. 303.It Fl u Ar dev 304Unconfigure the RAIDframe device. 305This does not remove any component labels or change any configuration 306settings (e.g. auto-configuration settings) for the RAID set. 307.It Fl v 308Be more verbose. 309For operations such as reconstructions, parity 310re-writing, and copybacks, provide a progress indicator. 311.El 312.Pp 313The device used by 314.Nm 315is specified by 316.Ar dev . 317.Ar dev 318may be either the full name of the device, e.g., 319.Pa /dev/rraid0d , 320for the i386 architecture, or 321.Pa /dev/rraid0c 322for many others, or just simply 323.Pa raid0 324(for 325.Pa /dev/rraid0[cd] ) . 326It is recommended that the partitions used to represent the 327RAID device are not used for file systems. 328.Ss Configuration file 329The format of the configuration file is complex, and 330only an abbreviated treatment is given here. 331In the configuration files, a 332.Sq # 333indicates the beginning of a comment. 334.Pp 335There are 4 required sections of a configuration file, and 2 336optional sections. 337Each section begins with a 338.Sq START , 339followed by the section name, 340and the configuration parameters associated with that section. 341The first section is the 342.Sq array 343section, and it specifies 344the number of rows, columns, and spare disks in the RAID set. 345For example: 346.Bd -literal -offset indent 347START array 3481 3 0 349.Ed 350.Pp 351indicates an array with 1 row, 3 columns, and 0 spare disks. 352Note that although multi-dimensional arrays may be specified, they are 353.Em NOT 354supported in the driver. 355.Pp 356The second section, the 357.Sq disks 358section, specifies the actual components of the device. 359For example: 360.Bd -literal -offset indent 361START disks 362/dev/sd0e 363/dev/sd1e 364/dev/sd2e 365.Ed 366.Pp 367specifies the three component disks to be used in the RAID device. 368If any of the specified drives cannot be found when the RAID device is 369configured, then they will be marked as 370.Sq failed , 371and the system will operate in degraded mode. 372Note that it is 373.Em imperative 374that the order of the components in the configuration file does not 375change between configurations of a RAID device. 376Changing the order of the components will result in data loss 377if the set is configured with the 378.Fl C 379option. 380In normal circumstances, the RAID set will not configure if only 381.Fl c 382is specified, and the components are out-of-order. 383.Pp 384The next section, which is the 385.Sq spare 386section, is optional, and, if present, specifies the devices to be used as 387.Sq hot spares 388\(em devices which are on-line, 389but are not actively used by the RAID driver unless 390one of the main components fail. 391A simple 392.Sq spare 393section might be: 394.Bd -literal -offset indent 395START spare 396/dev/sd3e 397.Ed 398.Pp 399for a configuration with a single spare component. 400If no spare drives are to be used in the configuration, then the 401.Sq spare 402section may be omitted. 403.Pp 404The next section is the 405.Sq layout 406section. 407This section describes the general layout parameters for the RAID device, 408and provides such information as 409sectors per stripe unit, 410stripe units per parity unit, 411stripe units per reconstruction unit, 412and the parity configuration to use. 413This section might look like: 414.Bd -literal -offset indent 415START layout 416# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level 41732 1 1 5 418.Ed 419.Pp 420The sectors per stripe unit specifies, in blocks, the interleave 421factor; i.e., the number of contiguous sectors to be written to each 422component for a single stripe. 423Appropriate selection of this value (32 in this example) 424is the subject of much research in RAID architectures. 425The stripe units per parity unit and 426stripe units per reconstruction unit are normally each set to 1. 427While certain values above 1 are permitted, a discussion of valid 428values and the consequences of using anything other than 1 are outside 429the scope of this document. 430The last value in this section (5 in this example) 431indicates the parity configuration desired. 432Valid entries include: 433.Bl -tag -width inde 434.It 0 435RAID level 0. 436No parity, only simple striping. 437.It 1 438RAID level 1. 439Mirroring. 440The parity is the mirror. 441.It 4 442RAID level 4. 443Striping across components, with parity stored on the last component. 444.It 5 445RAID level 5. 446Striping across components, parity distributed across all components. 447.El 448.Pp 449There are other valid entries here, including those for Even-Odd 450parity, RAID level 5 with rotated sparing, Chained declustering, 451and Interleaved declustering, but as of this writing the code for 452those parity operations has not been tested with 453.Nx . 454.Pp 455The next required section is the 456.Sq queue 457section. 458This is most often specified as: 459.Bd -literal -offset indent 460START queue 461fifo 100 462.Ed 463.Pp 464where the queuing method is specified as fifo (first-in, first-out), 465and the size of the per-component queue is limited to 100 requests. 466Other queuing methods may also be specified, but a discussion of them 467is beyond the scope of this document. 468.Pp 469The final section, the 470.Sq debug 471section, is optional. 472For more details on this the reader is referred to 473the RAIDframe documentation discussed in the 474.Sx HISTORY 475section. 476.Pp 477See 478.Sx EXAMPLES 479for a more complete configuration file example. 480.Sh FILES 481.Bl -tag -width /dev/XXrXraidX -compact 482.It Pa /dev/{,r}raid* 483.Cm raid 484device special files. 485.El 486.Sh EXAMPLES 487It is highly recommended that before using the RAID driver for real 488file systems that the system administrator(s) become quite familiar 489with the use of 490.Nm , 491and that they understand how the component reconstruction process works. 492The examples in this section will focus on configuring a 493number of different RAID sets of varying degrees of redundancy. 494By working through these examples, administrators should be able to 495develop a good feel for how to configure a RAID set, and how to 496initiate reconstruction of failed components. 497.Pp 498In the following examples 499.Sq raid0 500will be used to denote the RAID device. 501Depending on the architecture, 502.Pa /dev/rraid0c 503or 504.Pa /dev/rraid0d 505may be used in place of 506.Pa raid0 . 507.Ss Initialization and Configuration 508The initial step in configuring a RAID set is to identify the components 509that will be used in the RAID set. 510All components should be the same size. 511Each component should have a disklabel type of 512.Dv FS_RAID , 513and a typical disklabel entry for a RAID component might look like: 514.Bd -literal -offset indent 515f: 1800000 200495 RAID # (Cyl. 405*- 4041*) 516.Ed 517.Pp 518While 519.Dv FS_BSDFFS 520will also work as the component type, the type 521.Dv FS_RAID 522is preferred for RAIDframe use, as it is required for features such as 523auto-configuration. 524As part of the initial configuration of each RAID set, 525each component will be given a 526.Sq component label . 527A 528.Sq component label 529contains important information about the component, including a 530user-specified serial number, the row and column of that component in 531the RAID set, the redundancy level of the RAID set, a 532.Sq modification counter , 533and whether the parity information (if any) on that 534component is known to be correct. 535Component labels are an integral part of the RAID set, 536since they are used to ensure that components 537are configured in the correct order, and used to keep track of other 538vital information about the RAID set. 539Component labels are also required for the auto-detection 540and auto-configuration of RAID sets at boot time. 541For a component label to be considered valid, that 542particular component label must be in agreement with the other 543component labels in the set. 544For example, the serial number, 545.Sq modification counter , 546number of rows and number of columns must all be in agreement. 547If any of these are different, then the component is 548not considered to be part of the set. 549See 550.Xr raid 4 551for more information about component labels. 552.Pp 553Once the components have been identified, and the disks have 554appropriate labels, 555.Nm 556is then used to configure the 557.Xr raid 4 558device. 559To configure the device, a configuration file which looks something like: 560.Bd -literal -offset indent 561START array 562# numRow numCol numSpare 5631 3 1 564 565START disks 566/dev/sd1e 567/dev/sd2e 568/dev/sd3e 569 570START spare 571/dev/sd4e 572 573START layout 574# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_5 57532 1 1 5 576 577START queue 578fifo 100 579.Ed 580.Pp 581is created in a file. 582The above configuration file specifies a RAID 5 583set consisting of the components 584.Pa /dev/sd1e , 585.Pa /dev/sd2e , 586and 587.Pa /dev/sd3e , 588with 589.Pa /dev/sd4e 590available as a 591.Sq hot spare 592in case one of the three main drives should fail. 593A RAID 0 set would be specified in a similar way: 594.Bd -literal -offset indent 595START array 596# numRow numCol numSpare 5971 4 0 598 599START disks 600/dev/sd10e 601/dev/sd11e 602/dev/sd12e 603/dev/sd13e 604 605START layout 606# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_0 60764 1 1 0 608 609START queue 610fifo 100 611.Ed 612.Pp 613In this case, devices 614.Pa /dev/sd10e , 615.Pa /dev/sd11e , 616.Pa /dev/sd12e , 617and 618.Pa /dev/sd13e 619are the components that make up this RAID set. 620Note that there are no hot spares for a RAID 0 set, 621since there is no way to recover data if any of the components fail. 622.Pp 623For a RAID 1 (mirror) set, the following configuration might be used: 624.Bd -literal -offset indent 625START array 626# numRow numCol numSpare 6271 2 0 628 629START disks 630/dev/sd20e 631/dev/sd21e 632 633START layout 634# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_1 635128 1 1 1 636 637START queue 638fifo 100 639.Ed 640.Pp 641In this case, 642.Pa /dev/sd20e 643and 644.Pa /dev/sd21e 645are the two components of the mirror set. 646While no hot spares have been specified in this 647configuration, they easily could be, just as they were specified in 648the RAID 5 case above. 649Note as well that RAID 1 sets are currently limited to only 2 components. 650At present, n-way mirroring is not possible. 651.Pp 652The first time a RAID set is configured, the 653.Fl C 654option must be used: 655.Bd -literal -offset indent 656raidctl -C raid0.conf raid0 657.Ed 658.Pp 659where 660.Pa raid0.conf 661is the name of the RAID configuration file. 662The 663.Fl C 664forces the configuration to succeed, even if any of the component 665labels are incorrect. 666The 667.Fl C 668option should not be used lightly in 669situations other than initial configurations, as if 670the system is refusing to configure a RAID set, there is probably a 671very good reason for it. 672After the initial configuration is done (and 673appropriate component labels are added with the 674.Fl I 675option) then raid0 can be configured normally with: 676.Bd -literal -offset indent 677raidctl -c raid0.conf raid0 678.Ed 679.Pp 680When the RAID set is configured for the first time, it is 681necessary to initialize the component labels, and to initialize the 682parity on the RAID set. 683Initializing the component labels is done with: 684.Bd -literal -offset indent 685raidctl -I 112341 raid0 686.Ed 687.Pp 688where 689.Sq 112341 690is a user-specified serial number for the RAID set. 691This initialization step is 692.Em required 693for all RAID sets. 694As well, using different serial numbers between RAID sets is 695.Em strongly encouraged , 696as using the same serial number for all RAID sets will only serve to 697decrease the usefulness of the component label checking. 698.Pp 699Initializing the RAID set is done via the 700.Fl i 701option. 702This initialization 703.Em MUST 704be done for 705.Em all 706RAID sets, since among other things it verifies that the parity (if 707any) on the RAID set is correct. 708Since this initialization may be quite time-consuming, the 709.Fl v 710option may be also used in conjunction with 711.Fl i : 712.Bd -literal -offset indent 713raidctl -iv raid0 714.Ed 715.Pp 716This will give more verbose output on the 717status of the initialization: 718.Bd -literal -offset indent 719Initiating re-write of parity 720Parity Re-write status: 721 10% |**** | ETA: 06:03 / 722.Ed 723.Pp 724The output provides a 725.Sq Percent Complete 726in both a numeric and graphical format, as well as an estimated time 727to completion of the operation. 728.Pp 729Since it is the parity that provides the 730.Sq redundancy 731part of RAID, it is critical that the parity is correct as much as possible. 732If the parity is not correct, then there is no 733guarantee that data will not be lost if a component fails. 734.Pp 735Once the parity is known to be correct, it is then safe to perform 736.Xr disklabel 8 , 737.Xr newfs 8 , 738or 739.Xr fsck 8 740on the device or its file systems, and then to mount the file systems 741for use. 742.Pp 743Under certain circumstances (e.g., the additional component has not 744arrived, or data is being migrated off of a disk destined to become a 745component) it may be desirable to configure a RAID 1 set with only 746a single component. 747This can be achieved by using the word 748.Dq absent 749to indicate that a particular component is not present. 750In the following: 751.Bd -literal -offset indent 752START array 753# numRow numCol numSpare 7541 2 0 755 756START disks 757absent 758/dev/sd0e 759 760START layout 761# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_1 762128 1 1 1 763 764START queue 765fifo 100 766.Ed 767.Pp 768.Pa /dev/sd0e 769is the real component, and will be the second disk of a RAID 1 set. 770The first component is simply marked as being absent. 771Configuration (using 772.Fl C 773and 774.Fl I Ar 12345 775as above) proceeds normally, but initialization of the RAID set will 776have to wait until all physical components are present. 777After configuration, this set can be used normally, but will be operating 778in degraded mode. 779Once a second physical component is obtained, it can be hot-added, 780the existing data mirrored, and normal operation resumed. 781.Pp 782The size of the resulting RAID set will depend on the number of data 783components in the set. 784Space is automatically reserved for the component labels, and 785the actual amount of space used 786for data on a component will be rounded down to the largest possible 787multiple of the sectors per stripe unit (sectPerSU) value. 788Thus, the amount of space provided by the RAID set will be less 789than the sum of the size of the components. 790.Ss Maintenance of the RAID set 791After the parity has been initialized for the first time, the command: 792.Bd -literal -offset indent 793raidctl -p raid0 794.Ed 795.Pp 796can be used to check the current status of the parity. 797To check the parity and rebuild it necessary (for example, 798after an unclean shutdown) the command: 799.Bd -literal -offset indent 800raidctl -P raid0 801.Ed 802.Pp 803is used. 804Note that re-writing the parity can be done while 805other operations on the RAID set are taking place (e.g., while doing a 806.Xr fsck 8 807on a file system on the RAID set). 808However: for maximum effectiveness of the RAID set, the parity should be 809known to be correct before any data on the set is modified. 810.Pp 811To see how the RAID set is doing, the following command can be used to 812show the RAID set's status: 813.Bd -literal -offset indent 814raidctl -s raid0 815.Ed 816.Pp 817The output will look something like: 818.Bd -literal -offset indent 819Components: 820 /dev/sd1e: optimal 821 /dev/sd2e: optimal 822 /dev/sd3e: optimal 823Spares: 824 /dev/sd4e: spare 825Component label for /dev/sd1e: 826 Row: 0 Column: 0 Num Rows: 1 Num Columns: 3 827 Version: 2 Serial Number: 13432 Mod Counter: 65 828 Clean: No Status: 0 829 sectPerSU: 32 SUsPerPU: 1 SUsPerRU: 1 830 RAID Level: 5 blocksize: 512 numBlocks: 1799936 831 Autoconfig: No 832 Last configured as: raid0 833Component label for /dev/sd2e: 834 Row: 0 Column: 1 Num Rows: 1 Num Columns: 3 835 Version: 2 Serial Number: 13432 Mod Counter: 65 836 Clean: No Status: 0 837 sectPerSU: 32 SUsPerPU: 1 SUsPerRU: 1 838 RAID Level: 5 blocksize: 512 numBlocks: 1799936 839 Autoconfig: No 840 Last configured as: raid0 841Component label for /dev/sd3e: 842 Row: 0 Column: 2 Num Rows: 1 Num Columns: 3 843 Version: 2 Serial Number: 13432 Mod Counter: 65 844 Clean: No Status: 0 845 sectPerSU: 32 SUsPerPU: 1 SUsPerRU: 1 846 RAID Level: 5 blocksize: 512 numBlocks: 1799936 847 Autoconfig: No 848 Last configured as: raid0 849Parity status: clean 850Reconstruction is 100% complete. 851Parity Re-write is 100% complete. 852Copyback is 100% complete. 853.Ed 854.Pp 855This indicates that all is well with the RAID set. 856Of importance here are the component lines which read 857.Sq optimal , 858and the 859.Sq Parity status 860line. 861.Sq Parity status: clean 862indicates that the parity is up-to-date for this RAID set, 863whether or not the RAID set is in redundant or degraded mode. 864.Sq Parity status: DIRTY 865indicates that it is not known if the parity information is 866consistent with the data, and that the parity information needs 867to be checked. 868Note that if there are file systems open on the RAID set, 869the individual components will not be 870.Sq clean 871but the set as a whole can still be clean. 872.Pp 873To check the component label of 874.Pa /dev/sd1e , 875the following is used: 876.Bd -literal -offset indent 877raidctl -g /dev/sd1e raid0 878.Ed 879.Pp 880The output of this command will look something like: 881.Bd -literal -offset indent 882Component label for /dev/sd1e: 883 Row: 0 Column: 0 Num Rows: 1 Num Columns: 3 884 Version: 2 Serial Number: 13432 Mod Counter: 65 885 Clean: No Status: 0 886 sectPerSU: 32 SUsPerPU: 1 SUsPerRU: 1 887 RAID Level: 5 blocksize: 512 numBlocks: 1799936 888 Autoconfig: No 889 Last configured as: raid0 890.Ed 891.Ss Dealing with Component Failures 892If for some reason 893(perhaps to test reconstruction) it is necessary to pretend a drive 894has failed, the following will perform that function: 895.Bd -literal -offset indent 896raidctl -f /dev/sd2e raid0 897.Ed 898.Pp 899The system will then be performing all operations in degraded mode, 900where missing data is re-computed from existing data and the parity. 901In this case, obtaining the status of raid0 will return (in part): 902.Bd -literal -offset indent 903Components: 904 /dev/sd1e: optimal 905 /dev/sd2e: failed 906 /dev/sd3e: optimal 907Spares: 908 /dev/sd4e: spare 909.Ed 910.Pp 911Note that with the use of 912.Fl f 913a reconstruction has not been started. 914To both fail the disk and start a reconstruction, the 915.Fl F 916option must be used: 917.Bd -literal -offset indent 918raidctl -F /dev/sd2e raid0 919.Ed 920.Pp 921The 922.Fl f 923option may be used first, and then the 924.Fl F 925option used later, on the same disk, if desired. 926Immediately after the reconstruction is started, the status will report: 927.Bd -literal -offset indent 928Components: 929 /dev/sd1e: optimal 930 /dev/sd2e: reconstructing 931 /dev/sd3e: optimal 932Spares: 933 /dev/sd4e: used_spare 934[...] 935Parity status: clean 936Reconstruction is 10% complete. 937Parity Re-write is 100% complete. 938Copyback is 100% complete. 939.Ed 940.Pp 941This indicates that a reconstruction is in progress. 942To find out how the reconstruction is progressing the 943.Fl S 944option may be used. 945This will indicate the progress in terms of the 946percentage of the reconstruction that is completed. 947When the reconstruction is finished the 948.Fl s 949option will show: 950.Bd -literal -offset indent 951Components: 952 /dev/sd1e: optimal 953 /dev/sd2e: spared 954 /dev/sd3e: optimal 955Spares: 956 /dev/sd4e: used_spare 957[...] 958Parity status: clean 959Reconstruction is 100% complete. 960Parity Re-write is 100% complete. 961Copyback is 100% complete. 962.Ed 963.Pp 964At this point there are at least two options. 965First, if 966.Pa /dev/sd2e 967is known to be good (i.e., the failure was either caused by 968.Fl f 969or 970.Fl F , 971or the failed disk was replaced), then a copyback of the data can 972be initiated with the 973.Fl B 974option. 975In this example, this would copy the entire contents of 976.Pa /dev/sd4e 977to 978.Pa /dev/sd2e . 979Once the copyback procedure is complete, the 980status of the device would be (in part): 981.Bd -literal -offset indent 982Components: 983 /dev/sd1e: optimal 984 /dev/sd2e: optimal 985 /dev/sd3e: optimal 986Spares: 987 /dev/sd4e: spare 988.Ed 989.Pp 990and the system is back to normal operation. 991.Pp 992The second option after the reconstruction is to simply use 993.Pa /dev/sd4e 994in place of 995.Pa /dev/sd2e 996in the configuration file. 997For example, the configuration file (in part) might now look like: 998.Bd -literal -offset indent 999START array 10001 3 0 1001 1002START disks 1003/dev/sd1e 1004/dev/sd4e 1005/dev/sd3e 1006.Ed 1007.Pp 1008This can be done as 1009.Pa /dev/sd4e 1010is completely interchangeable with 1011.Pa /dev/sd2e 1012at this point. 1013Note that extreme care must be taken when 1014changing the order of the drives in a configuration. 1015This is one of the few instances where the devices and/or 1016their orderings can be changed without loss of data! 1017In general, the ordering of components in a configuration file should 1018.Em never 1019be changed. 1020.Pp 1021If a component fails and there are no hot spares 1022available on-line, the status of the RAID set might (in part) look like: 1023.Bd -literal -offset indent 1024Components: 1025 /dev/sd1e: optimal 1026 /dev/sd2e: failed 1027 /dev/sd3e: optimal 1028No spares. 1029.Ed 1030.Pp 1031In this case there are a number of options. 1032The first option is to add a hot spare using: 1033.Bd -literal -offset indent 1034raidctl -a /dev/sd4e raid0 1035.Ed 1036.Pp 1037After the hot add, the status would then be: 1038.Bd -literal -offset indent 1039Components: 1040 /dev/sd1e: optimal 1041 /dev/sd2e: failed 1042 /dev/sd3e: optimal 1043Spares: 1044 /dev/sd4e: spare 1045.Ed 1046.Pp 1047Reconstruction could then take place using 1048.Fl F 1049as describe above. 1050.Pp 1051A second option is to rebuild directly onto 1052.Pa /dev/sd2e . 1053Once the disk containing 1054.Pa /dev/sd2e 1055has been replaced, one can simply use: 1056.Bd -literal -offset indent 1057raidctl -R /dev/sd2e raid0 1058.Ed 1059.Pp 1060to rebuild the 1061.Pa /dev/sd2e 1062component. 1063As the rebuilding is in progress, the status will be: 1064.Bd -literal -offset indent 1065Components: 1066 /dev/sd1e: optimal 1067 /dev/sd2e: reconstructing 1068 /dev/sd3e: optimal 1069No spares. 1070.Ed 1071.Pp 1072and when completed, will be: 1073.Bd -literal -offset indent 1074Components: 1075 /dev/sd1e: optimal 1076 /dev/sd2e: optimal 1077 /dev/sd3e: optimal 1078No spares. 1079.Ed 1080.Pp 1081In circumstances where a particular component is completely 1082unavailable after a reboot, a special component name will be used to 1083indicate the missing component. 1084For example: 1085.Bd -literal -offset indent 1086Components: 1087 /dev/sd2e: optimal 1088 component1: failed 1089No spares. 1090.Ed 1091.Pp 1092indicates that the second component of this RAID set was not detected 1093at all by the auto-configuration code. 1094The name 1095.Sq component1 1096can be used anywhere a normal component name would be used. 1097For example, to add a hot spare to the above set, and rebuild to that hot 1098spare, the following could be done: 1099.Bd -literal -offset indent 1100raidctl -a /dev/sd3e raid0 1101raidctl -F component1 raid0 1102.Ed 1103.Pp 1104at which point the data missing from 1105.Sq component1 1106would be reconstructed onto 1107.Pa /dev/sd3e . 1108.Pp 1109When more than one component is marked as 1110.Sq failed 1111due to a non-component hardware failure (e.g., loss of power to two 1112components, adapter problems, termination problems, or cabling issues) it 1113is quite possible to recover the data on the RAID set. 1114The first thing to be aware of is that the first disk to fail will 1115almost certainly be out-of-sync with the remainder of the array. 1116If any IO was performed between the time the first component is considered 1117.Sq failed 1118and when the second component is considered 1119.Sq failed , 1120then the first component to fail will 1121.Em not 1122contain correct data, and should be ignored. 1123When the second component is marked as failed, however, the RAID device will 1124(currently) panic the system. 1125At this point the data on the RAID set 1126(not including the first failed component) is still self consistent, 1127and will be in no worse state of repair than had the power gone out in 1128the middle of a write to a file system on a non-RAID device. 1129The problem, however, is that the component labels may now have 3 different 1130.Sq modification counters 1131(one value on the first component that failed, one value on the second 1132component that failed, and a third value on the remaining components). 1133In such a situation, the RAID set will not autoconfigure, 1134and can only be forcibly re-configured 1135with the 1136.Fl C 1137option. 1138To recover the RAID set, one must first remedy whatever physical 1139problem caused the multiple-component failure. 1140After that is done, the RAID set can be restored by forcibly 1141configuring the raid set 1142.Em without 1143the component that failed first. 1144For example, if 1145.Pa /dev/sd1e 1146and 1147.Pa /dev/sd2e 1148fail (in that order) in a RAID set of the following configuration: 1149.Bd -literal -offset indent 1150START array 11511 4 0 1152 1153START disks 1154/dev/sd1e 1155/dev/sd2e 1156/dev/sd3e 1157/dev/sd4e 1158 1159START layout 1160# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_5 116164 1 1 5 1162 1163START queue 1164fifo 100 1165 1166.Ed 1167.Pp 1168then the following configuration (say "recover_raid0.conf") 1169.Bd -literal -offset indent 1170START array 11711 4 0 1172 1173START disks 1174absent 1175/dev/sd2e 1176/dev/sd3e 1177/dev/sd4e 1178 1179START layout 1180# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_5 118164 1 1 5 1182 1183START queue 1184fifo 100 1185.Ed 1186.Pp 1187can be used with 1188.Bd -literal -offset indent 1189raidctl -C recover_raid0.conf raid0 1190.Ed 1191.Pp 1192to force the configuration of raid0. 1193A 1194.Bd -literal -offset indent 1195raidctl -I 12345 raid0 1196.Ed 1197.Pp 1198will be required in order to synchronize the component labels. 1199At this point the file systems on the RAID set can then be checked and 1200corrected. 1201To complete the re-construction of the RAID set, 1202.Pa /dev/sd1e 1203is simply hot-added back into the array, and reconstructed 1204as described earlier. 1205.Ss RAID on RAID 1206RAID sets can be layered to create more complex and much larger RAID sets. 1207A RAID 0 set, for example, could be constructed from four RAID 5 sets. 1208The following configuration file shows such a setup: 1209.Bd -literal -offset indent 1210START array 1211# numRow numCol numSpare 12121 4 0 1213 1214START disks 1215/dev/raid1e 1216/dev/raid2e 1217/dev/raid3e 1218/dev/raid4e 1219 1220START layout 1221# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_0 1222128 1 1 0 1223 1224START queue 1225fifo 100 1226.Ed 1227.Pp 1228A similar configuration file might be used for a RAID 0 set 1229constructed from components on RAID 1 sets. 1230In such a configuration, the mirroring provides a high degree 1231of redundancy, while the striping provides additional speed benefits. 1232.Ss Auto-configuration and Root on RAID 1233RAID sets can also be auto-configured at boot. 1234To make a set auto-configurable, 1235simply prepare the RAID set as above, and then do a: 1236.Bd -literal -offset indent 1237raidctl -A yes raid0 1238.Ed 1239.Pp 1240to turn on auto-configuration for that set. 1241To turn off auto-configuration, use: 1242.Bd -literal -offset indent 1243raidctl -A no raid0 1244.Ed 1245.Pp 1246RAID sets which are auto-configurable will be configured before the 1247root file system is mounted. 1248These RAID sets are thus available for 1249use as a root file system, or for any other file system. 1250A primary advantage of using the auto-configuration is that RAID components 1251become more independent of the disks they reside on. 1252For example, SCSI ID's can change, but auto-configured sets will always be 1253configured correctly, even if the SCSI ID's of the component disks 1254have become scrambled. 1255.Pp 1256Having a system's root file system 1257.Pq Pa / 1258on a RAID set is also allowed, with the 1259.Sq a 1260partition of such a RAID set being used for 1261.Pa / . 1262To use raid0a as the root file system, simply use: 1263.Bd -literal -offset indent 1264raidctl -A root raid0 1265.Ed 1266.Pp 1267To return raid0a to be just an auto-configuring set simply use the 1268.Fl A Ar yes 1269arguments. 1270.Pp 1271Note that kernels can only be directly read from RAID 1 components on 1272architectures that support that 1273.Pq currently alpha, i386, pmax, sparc, sparc64, and vax . 1274On those architectures, the 1275.Dv FS_RAID 1276file system is recognized by the bootblocks, and will properly load the 1277kernel directly from a RAID 1 component. 1278For other architectures, or to support the root file system 1279on other RAID sets, some other mechanism must be used to get a kernel booting. 1280For example, a small partition containing only the secondary boot-blocks 1281and an alternate kernel (or two) could be used. 1282Once a kernel is booting however, and an auto-configuring RAID set is 1283found that is eligible to be root, then that RAID set will be 1284auto-configured and used as the root device. 1285If two or more RAID sets claim to be root devices, then the 1286user will be prompted to select the root device. 1287At this time, RAID 0, 1, 4, and 5 sets are all supported as root devices. 1288.Pp 1289A typical RAID 1 setup with root on RAID might be as follows: 1290.Bl -enum 1291.It 1292wd0a - a small partition, which contains a complete, bootable, basic 1293.Nx 1294installation. 1295.It 1296wd1a - also contains a complete, bootable, basic 1297.Nx 1298installation. 1299.It 1300wd0e and wd1e - a RAID 1 set, raid0, used for the root file system. 1301.It 1302wd0f and wd1f - a RAID 1 set, raid1, which will be used only for 1303swap space. 1304.It 1305wd0g and wd1g - a RAID 1 set, raid2, used for 1306.Pa /usr , 1307.Pa /home , 1308or other data, if desired. 1309.It 1310wd0h and wd1h - a RAID 1 set, raid3, if desired. 1311.El 1312.Pp 1313RAID sets raid0, raid1, and raid2 are all marked as auto-configurable. 1314raid0 is marked as being a root file system. 1315When new kernels are installed, the kernel is not only copied to 1316.Pa / , 1317but also to wd0a and wd1a. 1318The kernel on wd0a is required, since that 1319is the kernel the system boots from. 1320The kernel on wd1a is also 1321required, since that will be the kernel used should wd0 fail. 1322The important point here is to have redundant copies of the kernel 1323available, in the event that one of the drives fail. 1324.Pp 1325There is no requirement that the root file system be on the same disk 1326as the kernel. 1327For example, obtaining the kernel from wd0a, and using 1328sd0e and sd1e for raid0, and the root file system, is fine. 1329It 1330.Em is 1331critical, however, that there be multiple kernels available, in the 1332event of media failure. 1333.Pp 1334Multi-layered RAID devices (such as a RAID 0 set made 1335up of RAID 1 sets) are 1336.Em not 1337supported as root devices or auto-configurable devices at this point. 1338(Multi-layered RAID devices 1339.Em are 1340supported in general, however, as mentioned earlier.) 1341Note that in order to enable component auto-detection and 1342auto-configuration of RAID devices, the line: 1343.Bd -literal -offset indent 1344options RAID_AUTOCONFIG 1345.Ed 1346.Pp 1347must be in the kernel configuration file. 1348See 1349.Xr raid 4 1350for more details. 1351.Ss Swapping on RAID 1352A RAID device can be used as a swap device. 1353In order to ensure that a RAID device used as a swap device 1354is correctly unconfigured when the system is shutdown or rebooted, 1355it is recommended that the line 1356.Bd -literal -offset indent 1357swapoff=YES 1358.Ed 1359.Pp 1360be added to 1361.Pa /etc/rc.conf . 1362.Ss Unconfiguration 1363The final operation performed by 1364.Nm 1365is to unconfigure a 1366.Xr raid 4 1367device. 1368This is accomplished via a simple: 1369.Bd -literal -offset indent 1370raidctl -u raid0 1371.Ed 1372.Pp 1373at which point the device is ready to be reconfigured. 1374.Ss Performance Tuning 1375Selection of the various parameter values which result in the best 1376performance can be quite tricky, and often requires a bit of 1377trial-and-error to get those values most appropriate for a given system. 1378A whole range of factors come into play, including: 1379.Bl -enum 1380.It 1381Types of components (e.g., SCSI vs. IDE) and their bandwidth 1382.It 1383Types of controller cards and their bandwidth 1384.It 1385Distribution of components among controllers 1386.It 1387IO bandwidth 1388.It 1389file system access patterns 1390.It 1391CPU speed 1392.El 1393.Pp 1394As with most performance tuning, benchmarking under real-life loads 1395may be the only way to measure expected performance. 1396Understanding some of the underlying technology is also useful in tuning. 1397The goal of this section is to provide pointers to those parameters which may 1398make significant differences in performance. 1399.Pp 1400For a RAID 1 set, a SectPerSU value of 64 or 128 is typically sufficient. 1401Since data in a RAID 1 set is arranged in a linear 1402fashion on each component, selecting an appropriate stripe size is 1403somewhat less critical than it is for a RAID 5 set. 1404However: a stripe size that is too small will cause large IO's to be 1405broken up into a number of smaller ones, hurting performance. 1406At the same time, a large stripe size may cause problems with 1407concurrent accesses to stripes, which may also affect performance. 1408Thus values in the range of 32 to 128 are often the most effective. 1409.Pp 1410Tuning RAID 5 sets is trickier. 1411In the best case, IO is presented to the RAID set one stripe at a time. 1412Since the entire stripe is available at the beginning of the IO, 1413the parity of that stripe can be calculated before the stripe is written, 1414and then the stripe data and parity can be written in parallel. 1415When the amount of data being written is less than a full stripe worth, the 1416.Sq small write 1417problem occurs. 1418Since a 1419.Sq small write 1420means only a portion of the stripe on the components is going to 1421change, the data (and parity) on the components must be updated 1422slightly differently. 1423First, the 1424.Sq old parity 1425and 1426.Sq old data 1427must be read from the components. 1428Then the new parity is constructed, 1429using the new data to be written, and the old data and old parity. 1430Finally, the new data and new parity are written. 1431All this extra data shuffling results in a serious loss of performance, 1432and is typically 2 to 4 times slower than a full stripe write (or read). 1433To combat this problem in the real world, it may be useful 1434to ensure that stripe sizes are small enough that a 1435.Sq large IO 1436from the system will use exactly one large stripe write. 1437As is seen later, there are some file system dependencies 1438which may come into play here as well. 1439.Pp 1440Since the size of a 1441.Sq large IO 1442is often (currently) only 32K or 64K, on a 5-drive RAID 5 set it may 1443be desirable to select a SectPerSU value of 16 blocks (8K) or 32 1444blocks (16K). 1445Since there are 4 data sectors per stripe, the maximum 1446data per stripe is 64 blocks (32K) or 128 blocks (64K). 1447Again, empirical measurement will provide the best indicators of which 1448values will yield better performance. 1449.Pp 1450The parameters used for the file system are also critical to good performance. 1451For 1452.Xr newfs 8 , 1453for example, increasing the block size to 32K or 64K may improve 1454performance dramatically. 1455As well, changing the cylinders-per-group 1456parameter from 16 to 32 or higher is often not only necessary for 1457larger file systems, but may also have positive performance implications. 1458.Ss Summary 1459Despite the length of this man-page, configuring a RAID set is a 1460relatively straight-forward process. 1461All that needs to be done is the following steps: 1462.Bl -enum 1463.It 1464Use 1465.Xr disklabel 8 1466to create the components (of type RAID). 1467.It 1468Construct a RAID configuration file: e.g., 1469.Pa raid0.conf 1470.It 1471Configure the RAID set with: 1472.Bd -literal -offset indent 1473raidctl -C raid0.conf raid0 1474.Ed 1475.Pp 1476.It 1477Initialize the component labels with: 1478.Bd -literal -offset indent 1479raidctl -I 123456 raid0 1480.Ed 1481.Pp 1482.It 1483Initialize other important parts of the set with: 1484.Bd -literal -offset indent 1485raidctl -i raid0 1486.Ed 1487.Pp 1488.It 1489Get the default label for the RAID set: 1490.Bd -literal -offset indent 1491disklabel raid0 \*[Gt] /tmp/label 1492.Ed 1493.Pp 1494.It 1495Edit the label: 1496.Bd -literal -offset indent 1497vi /tmp/label 1498.Ed 1499.Pp 1500.It 1501Put the new label on the RAID set: 1502.Bd -literal -offset indent 1503disklabel -R -r raid0 /tmp/label 1504.Ed 1505.Pp 1506.It 1507Create the file system: 1508.Bd -literal -offset indent 1509newfs /dev/rraid0e 1510.Ed 1511.Pp 1512.It 1513Mount the file system: 1514.Bd -literal -offset indent 1515mount /dev/raid0e /mnt 1516.Ed 1517.Pp 1518.It 1519Use: 1520.Bd -literal -offset indent 1521raidctl -c raid0.conf raid0 1522.Ed 1523.Pp 1524To re-configure the RAID set the next time it is needed, or put 1525.Pa raid0.conf 1526into 1527.Pa /etc 1528where it will automatically be started by the 1529.Pa /etc/rc.d 1530scripts. 1531.El 1532.Sh SEE ALSO 1533.Xr ccd 4 , 1534.Xr raid 4 , 1535.Xr rc 8 1536.Sh HISTORY 1537RAIDframe is a framework for rapid prototyping of RAID structures 1538developed by the folks at the Parallel Data Laboratory at Carnegie 1539Mellon University (CMU). 1540A more complete description of the internals and functionality of 1541RAIDframe is found in the paper "RAIDframe: A Rapid Prototyping Tool 1542for RAID Systems", by William V. Courtright II, Garth Gibson, Mark 1543Holland, LeAnn Neal Reilly, and Jim Zelenka, and published by the 1544Parallel Data Laboratory of Carnegie Mellon University. 1545.Pp 1546The 1547.Nm 1548command first appeared as a program in CMU's RAIDframe v1.1 distribution. 1549This version of 1550.Nm 1551is a complete re-write, and first appeared in 1552.Nx 1.4 . 1553.Sh COPYRIGHT 1554.Bd -literal 1555The RAIDframe Copyright is as follows: 1556 1557Copyright (c) 1994-1996 Carnegie-Mellon University. 1558All rights reserved. 1559 1560Permission to use, copy, modify and distribute this software and 1561its documentation is hereby granted, provided that both the copyright 1562notice and this permission notice appear in all copies of the 1563software, derivative works or modified versions, and any portions 1564thereof, and that both notices appear in supporting documentation. 1565 1566CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" 1567CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND 1568FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. 1569 1570Carnegie Mellon requests users of this software to return to 1571 1572 Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU 1573 School of Computer Science 1574 Carnegie Mellon University 1575 Pittsburgh PA 15213-3890 1576 1577any improvements or extensions that they make and grant Carnegie the 1578rights to redistribute these changes. 1579.Ed 1580.Sh WARNINGS 1581Certain RAID levels (1, 4, 5, 6, and others) can protect against some 1582data loss due to component failure. 1583However the loss of two components of a RAID 4 or 5 system, 1584or the loss of a single component of a RAID 0 system will 1585result in the entire file system being lost. 1586RAID is 1587.Em NOT 1588a substitute for good backup practices. 1589.Pp 1590Recomputation of parity 1591.Em MUST 1592be performed whenever there is a chance that it may have been compromised. 1593This includes after system crashes, or before a RAID 1594device has been used for the first time. 1595Failure to keep parity correct will be catastrophic should a 1596component ever fail \(em it is better to use RAID 0 and get the 1597additional space and speed, than it is to use parity, but 1598not keep the parity correct. 1599At least with RAID 0 there is no perception of increased data security. 1600.Sh BUGS 1601Hot-spare removal is currently not available. 1602