Diff for /wikisrc/guide/raidframe.mdwn between versions 1.4 and 1.14

version 1.4, 2013/03/02 13:13:34 version 1.14, 2021/04/12 13:15:03
Line 1 Line 1
 # NetBSD RAIDframe  This page was moved to:
   [The NetBSD Guide - NetBSD RAIDframe](//www.NetBSD.org/docs/guide/en/chap-raidframe.html)
 ## RAIDframe Introduction  
   
 ### About RAIDframe  
   
 NetBSD uses the [CMU RAIDframe](http://www.pdl.cmu.edu/RAIDframe/) software for   
 its RAID subsystem. NetBSD is the primary platform for RAIDframe development.   
 RAIDframe can also be found in older versions of FreeBSD and OpenBSD. NetBSD   
 also has another way of bundling disks, the   
 [ccd(4)](http://netbsd.gw.com/cgi-bin/man-cgi?ccd+4+NetBSD-5.0.1+i386) subsystem   
 (see [Concatenated Disk Device](/guide/ccd)). You should possess some [basic   
 knowledge](http://www.acnc.com/04_00.html) about RAID concepts and terminology   
 before continuing. You should also be at least familiar with the different   
 levels of RAID - Adaptec provides an [excellent   
 reference](http://www.adaptec.com/en-US/_common/compatibility/_education/RAID_level_compar_wp.htm),   
 and the [raid(4)](http://netbsd.gw.com/cgi-bin/man-cgi?raid+4+NetBSD-5.0.1+i386)   
 manpage contains a short overview too.  
   
 ### A warning about Data Integrity, Backups, and High Availability  
   
 RAIDframe is a Software RAID implementation, as opposed to Hardware RAID. As   
 such, it does not need special disk controllers supported by NetBSD. System   
 administrators should give a great deal of consideration to whether software   
 RAID or hardware RAID is more appropriate for their "Mission Critical"   
 applications. For some projects you might consider the use of many of the   
 hardware RAID devices [supported by   
 NetBSD](http://www.NetBSD.org/support/hardware/). It is truly at your discretion   
 what type of RAID you use, but it is recommend that you consider factors such   
 as: manageability, commercial vendor support, load-balancing and failover, etc.  
   
 Depending on the RAID level used, RAIDframe does provide redundancy in the event   
 of a hardware failure. However, it is *not* a replacement for reliable backups!   
 Software and user-error can still cause data loss. RAIDframe may be used as a   
 mechanism for facilitating backups in systems without backup hardware, but this   
 is not an ideal configuration. Finally, with regard to "high availability", RAID   
 is only a very small component to ensuring data availability.  
   
 Once more for good measure: *Back up your data!*  
   
 ### Hardware versus Software RAID  
   
 If you run a server, it will most probably already have a Hardware RAID   
 controller. There are reasons for and against using a Software RAID, depending   
 on the scenario.  
   
 In general, a Software RAID is well suited for low-IO system disks. If you run a   
 Software RAID, you can exchange disks and disk controllers, or even move the   
 disks to a completely different machine. The computational overhead for the RAID   
 is negligible if there is only few disk IO operations.  
   
 If you need much IO, you should use a Hardware RAID. With a Software RAID, the   
 redundancy data has to be transferred via the bus your disk controller is   
 connected to. With a Hardware RAID, you transfer data only once - the redundancy   
 computation and transfer is done by the controller.  
   
 ### Getting Help  
   
 If you encounter problems using RAIDframe, you have several options for   
 obtaining help.  
   
  1. Read the RAIDframe man pages:   
     [raid(4)](http://netbsd.gw.com/cgi-bin/man-cgi?raid+4+NetBSD-5.0.1+i386) and   
     [raidctl(8)](http://netbsd.gw.com/cgi-bin/man-cgi?raidctl+8+NetBSD-5.0.1+i386)   
     thoroughly.  
   
  2. Search the mailing list archives. Unfortunately, there is no NetBSD list   
     dedicated to RAIDframe support. Depending on the nature of the problem, posts  
     tend to end up in a variety of lists. At a very minimum, search  
     [netbsd-help](http://mail-index.NetBSD.org/netbsd-help/),  
     [netbsd-users@NetBSD.org](http://mail-index.NetBSD.org/netbsd-users/),  
     [current-users@NetBSD.org](http://mail-index.NetBSD.org/current-users/). Also  
     search the list for the NetBSD platform on which you are using RAIDframe:  
     port-*`${ARCH}`*@NetBSD.org.  
   
     ### Caution  
   
         Because RAIDframe is constantly undergoing development, some information in   
         mailing list archives has the potential of being dated and inaccurate.  
   
  3. Search the [Problem Report   
     database](http://www.NetBSD.org/support/send-pr.html).  
   
  4. If your problem persists: Post to the mailing list most appropriate   
     (judgment call). Collect as much verbosely detailed information as possible   
     before posting: Include your   
     [dmesg(8)](http://netbsd.gw.com/cgi-bin/man-cgi?dmesg+8+NetBSD-5.0.1+i386)   
     output from `/var/run/dmesg.boot`, your kernel   
     [config(5)](http://netbsd.gw.com/cgi-bin/man-cgi?config+5+NetBSD-5.0.1+i386)  
     your `/etc/raid[0-9].conf`, any relevant errors on `/dev/console`,   
     `/var/log/messages`, or to `stdout/stderr` of   
     [raidctl(8)](http://netbsd.gw.com/cgi-bin/man-cgi?raidctl+8+NetBSD-5.0.1+i386).   
     The output of **raidctl -s** (if available) will be useful as well. Also   
     include details on the troubleshooting steps you've taken thus far, exactly   
     when the problem started, and any notes on recent changes that may have   
     prompted the problem to develop. Remember to be patient when waiting for a   
     response.  
   
 ## Setup RAIDframe Support  
   
 The use of RAID will require software and hardware configuration changes.  
   
 ### Kernel Support  
   
 The GENERIC kernel already has support for RAIDframe. If you have built a custom   
 kernel for your environment the kernel configuration must have the following   
 options:  
   
     pseudo-device   raid            8       # RAIDframe disk driver  
     options         RAID_AUTOCONFIG         # auto-configuration of RAID components  
   
 The RAID support must be detected by the NetBSD kernel, which can be checked by   
 looking at the output of the   
 [dmesg(8)](http://netbsd.gw.com/cgi-bin/man-cgi?dmesg+8+NetBSD-5.0.1+i386)   
 command.  
   
     # dmesg|grep -i raid  
     Kernelized RAIDframe activated  
   
 Historically, the kernel must also contain static mappings between bus addresses   
 and device nodes in `/dev`. This used to ensure consistency of devices within   
 RAID sets in the event of a device failure after reboot. Since NetBSD 1.6,   
 however, using the auto-configuration features of RAIDframe has been recommended   
 over statically mapping devices. The auto-configuration features allow drives to   
 move around on the system, and RAIDframe will automatically determine which   
 components belong to which RAID sets.  
   
 ### Power Redundancy and Disk Caching  
   
 If your system has an Uninterruptible Power Supply (UPS), if your system has   
 redundant power supplies, or your disk controller has a battery, you should   
 consider enabling the read and write caches on your drives. On systems with   
 redundant power, this will improve drive performance. On systems without   
 redundant power, the write cache could endanger the integrity of RAID data in   
 the event of a power loss.  
   
 The [dkctl(8)](http://netbsd.gw.com/cgi-bin/man-cgi?dkctl+8+NetBSD-5.0.1+i386)   
 utility to can be used for this on all kinds of disks that support the operation   
 (SCSI, EIDE, SATA, ...):  
   
     # dkctl wd0 getcache  
     /dev/rwd0d: read cache enabled  
     /dev/rwd0d: read cache enable is not changeable  
     /dev/rwd0d: write cache enable is changeable  
     /dev/rwd0d: cache parameters are not savable  
     # dkctl wd0 setcache rw  
     # dkctl wd0 getcache  
     /dev/rwd0d: read cache enabled  
     /dev/rwd0d: write-back cache enabled  
     /dev/rwd0d: read cache enable is not changeable  
     /dev/rwd0d: write cache enable is changeable  
     /dev/rwd0d: cache parameters are not savable  
   
 ## Example: RAID-1 Root Disk  
   
 This example explains how to setup RAID-1 root disk. With RAID-1 components are   
 mirrored and therefore the server can be fully functional in the event of a   
 single component failure. The goal is to provide a level of redundancy that will   
 allow the system to encounter a component failure on either component disk in   
 the RAID and:  
   
  * Continue normal operations until a maintenance window can be scheduled.  
  * Or, in the unlikely event that the component failure causes a system reboot,   
    be able to quickly reconfigure the system to boot from the remaining  
    component (platform dependent).  
   
 ![RAID-1 Disk Logical Layout](/guide/images/raidframe_raidL1-diskdia.png)  
 **RAID-1 Disk Logical Layout**  
   
 Because RAID-1 provides both redundancy and performance improvements, its most   
 practical application is on critical "system" partitions such as `/`, `/usr`,   
 `/var`, `swap`, etc., where read operations are more frequent than write   
 operations. For other file systems, such as `/home` or `/var/`, other RAID   
 levels might be considered (see the references above). If one were simply   
 creating a generic RAID-1 volume for a non-root file system, the cookie-cutter   
 examples from the man page could be followed, but because the root volume must   
 be bootable, certain special steps must be taken during initial setup.  
   
 *Note*: This example will outline a process that differs only slightly between   
 the i386 and sparc64 platforms. In an attempt to reduce excessive duplication of   
 content, where differences do exist and are cosmetic in nature, they will be   
 pointed out using a section such as this. If the process is drastically   
 different, the process will branch into separate, platform dependent steps.  
   
 ### Pseudo-Process Outline  
   
 Although a much more refined process could be developed using a custom copy of   
 NetBSD installed on custom-developed removable media, presently the NetBSD   
 install media lacks RAIDframe tools and support, so the following pseudo process  
 has become the de facto standard for setting up RAID-1 Root.  
   
  1. Install a stock NetBSD onto Disk0 of your system.  
   
     ![Perform generic install onto Disk0/wd0](/guide/images/raidframe_r1r-pp1.png)  
     **Perform generic install onto Disk0/wd0**  
   
  2. Use the installed system on Disk0/wd0 to setup a RAID Set composed of   
     Disk1/wd1 only.  
   
     ![Setup RAID Set](raidframe_r1r-pp2.png)  
     **Setup RAID Set**  
   
  3. Reboot the system off the Disk1/wd1 with the newly created RAID volume.  
   
     ![Reboot using Disk1/wd1 of RAID](/guide/images/raidframe_r1r-pp3.png)  
     **Reboot using Disk1/wd1 of RAID**  
   
  4. Add / re-sync Disk0/wd0 back into the RAID set.  
   
     ![Mirror Disk1/wd1 back to Disk0/wd0](/guide/images/raidframe_r1r-pp4.png)  
     **Mirror Disk1/wd1 back to Disk0/wd0**  
   
 ### Hardware Review  
   
 At present, the alpha, amd64, i386, pmax, sparc, sparc64, and vax NetBSD   
 platforms support booting from RAID-1. Booting is not supported from any other   
 RAID level. Booting from a RAID set is accomplished by teaching the 1st stage   
 boot loader to understand both 4.2BSD/FFS and RAID partitions. The 1st boot   
 block code only needs to know enough about the disk partitions and file systems   
 to be able to read the 2nd stage boot blocks. Therefore, at any time, the   
 system's BIOS / firmware must be able to read a drive with 1st stage boot blocks   
 installed. On the i386 platform, configuring this is entirely dependent on the   
 vendor of the controller card / host bus adapter to which your disks are   
 connected. On sparc64 this is controlled by the IEEE 1275 Sun OpenBoot Firmware.  
   
 This article assumes two identical IDE disks (`/dev/wd{0,1}`) which we are going   
 to mirror (RAID-1). These disks are identified as:  
   
     # grep ^wd /var/run/dmesg.boot  
     wd0 at atabus0 drive 0: <WDC WD100BB-75CLB0>  
     wd0: drive supports 16-sector PIO transfers, LBA addressing  
     wd0: 9541 MB, 19386 cyl, 16 head, 63 sec, 512 bytes/sect x 19541088 sectors  
     wd0: drive supports PIO mode 4, DMA mode 2, Ultra-DMA mode 5 (Ultra/100)  
     wd0(piixide0:0:0): using PIO mode 4, Ultra-DMA mode 2 (Ultra/33) (using DMA data transfers)  
       
     wd1 at atabus1 drive 0: <WDC WD100BB-75CLB0>  
     wd1: drive supports 16-sector PIO transfers, LBA addressing  
     wd1: 9541 MB, 19386 cyl, 16 head, 63 sec, 512 bytes/sect x 19541088 sectors  
     wd1: drive supports PIO mode 4, DMA mode 2, Ultra-DMA mode 5 (Ultra/100)  
     wd1(piixide0:1:0): using PIO mode 4, Ultra-DMA mode 2 (Ultra/33) (using DMA data transfers)  
   
 *Note*: If you are using SCSI, replace `/dev/{,r}wd{0,1}` with   
 `/dev/{,r}sd{0,1}`.  
   
 In this example, both disks are jumpered as Master on separate channels on the   
 same controller. You usually wouldn't want to have both disks on the same bus on   
 the same controller; this creates a single point of failure. Ideally you would   
 have the disks on separate channels on separate controllers. Nonetheless, in   
 most cases the most critical point is the hard disk, so having redundant   
 channels or controllers is not that important. Plus, having more channels or   
 controllers increases costs. Some SCSI controllers have multiple channels on the   
 same controller, however, a SCSI bus reset on one channel could adversely affect   
 the other channel if the ASIC/IC becomes overloaded. The trade-off with two   
 controllers is that twice the bandwidth is used on the system bus. For purposes   
 of simplification, this example shows two disks on different channels on the   
 same controller.  
   
 *Note*: RAIDframe requires that all components be of the same size. Actually, it   
 will use the lowest common denominator among components of dissimilar sizes. For   
 purposes of illustration, the example uses two disks of identical geometries.   
 Also, consider the availability of replacement disks if a component suffers a   
 critical hardware failure.  
   
 *Tip*: Two disks of identical vendor model numbers could have different   
 geometries if the drive possesses "grown defects". Use a low-level program to   
 examine the grown defects table of the disk. These disks are obviously   
 suboptimal candidates for use in RAID and should be avoided.  
   
 ### Initial Install on Disk0/wd0  
   
 Perform a very generic installation onto your Disk0/wd0. Follow the `INSTALL`   
 instructions for your platform. Install all the sets but do not bother   
 customizing anything other than the kernel as it will be overwritten.  
   
 *Tip*: On i386, during the sysinst install, when prompted if you want to `use   
 the entire disk for NetBSD`, answer `yes`.  
   
  * [Installing NetBSD: Preliminary considerations and preparations](/guide/inst)  
  * [NetBSD/i386 Install Directions](http://ftp.NetBSD.org/pub/NetBSD/NetBSD-5.0.2/i386/INSTALL.html)  
  * [NetBSD/sparc64 Install Directions](http://ftp.NetBSD.org/pub/NetBSD/NetBSD-5.0.2/sparc64/INSTALL.html)  
   
 Once the installation is complete, you should examine the   
 [disklabel(8)](http://netbsd.gw.com/cgi-bin/man-cgi?disklabel+8+NetBSD-5.0.1+i386)   
 and [fdisk(8)](http://netbsd.gw.com/cgi-bin/man-cgi?fdisk+8+NetBSD-5.0.1+i386)  
 [sunlabel(8)](http://netbsd.gw.com/cgi-bin/man-cgi?sunlabel+8+NetBSD-5.0.1+i386)   
 outputs on the system:  
   
     # df  
     Filesystem   1K-blocks        Used       Avail %Cap Mounted on  
     /dev/wd0a       9487886      502132     8511360   5% /  
   
 On i386:  
   
     # disklabel -r wd0  
     type: unknown  
     disk: Disk00  
     label:  
     flags:  
     bytes/sector: 512  
     sectors/track: 63  
     tracks/cylinder: 16  
     sectors/cylinder: 1008  
     cylinders: 19386  
     total sectors: 19541088  
     rpm: 3600  
     interleave: 1  
     trackskew: 0  
     cylinderskew: 0  
     headswitch: 0           # microseconds  
     track-to-track seek: 0  # microseconds  
     drivedata: 0  
       
     16 partitions:  
     #        size    offset     fstype [fsize bsize cpg/sgs]  
      a:  19276992        63     4.2BSD   1024  8192 46568  # (Cyl.      0* - 19124*)  
      b:    264033  19277055       swap                     # (Cyl.  19124* - 19385)  
      c:  19541025        63     unused      0     0        # (Cyl.      0* - 19385)  
      d:  19541088         0     unused      0     0        # (Cyl.      0 - 19385)  
       
     # fdisk /dev/rwd0d  
     Disk: /dev/rwd0d  
     NetBSD disklabel disk geometry:  
     cylinders: 19386, heads: 16, sectors/track: 63 (1008 sectors/cylinder)  
     total sectors: 19541088  
       
     BIOS disk geometry:  
     cylinders: 1023, heads: 255, sectors/track: 63 (16065 sectors/cylinder)  
     total sectors: 19541088  
       
     Partition table:  
     0: NetBSD (sysid 169)  
         start 63, size 19541025 (9542 MB, Cyls 0-1216/96/1), Active  
     1: <UNUSED>  
     2: <UNUSED>  
     3: <UNUSED>  
     Bootselector disabled.  
     First active partition: 0  
   
 On Sparc64 the command and output differ slightly:  
   
     # disklabel -r wd0  
     type: unknown  
     disk: Disk0  
     [...snip...]  
     8 partitions:  
     #        size    offset     fstype [fsize bsize cpg/sgs]  
      a:  19278000         0     4.2BSD   1024  8192 46568  # (Cyl.      0 -  19124)  
      b:    263088  19278000       swap                     # (Cyl.  19125 -  19385)  
      c:  19541088         0     unused      0     0        # (Cyl.      0 -  19385)  
       
     # sunlabel /dev/rwd0c  
     sunlabel> P  
     a: start cyl =      0, size = 19278000 (19125/0/0 - 9413.09Mb)  
     b: start cyl =  19125, size =   263088 (261/0/0 - 128.461Mb)  
     c: start cyl =      0, size = 19541088 (19386/0/0 - 9541.55Mb)  
   
 ### Preparing Disk1/wd1  
   
 Once you have a stock install of NetBSD on Disk0/wd0, you are ready to begin.   
 Disk1/wd1 will be visible and unused by the system. To setup Disk1/wd1, you will   
 use   
 [disklabel(8)](http://netbsd.gw.com/cgi-bin/man-cgi?disklabel+8+NetBSD-5.0.1+i386)   
 to allocate the entire second disk to the RAID-1 set.  
   
 *Tip*: The best way to ensure that Disk1/wd1 is completely empty is to 'zero'   
 out the first few sectors of the disk with   
 [dd(1)](http://netbsd.gw.com/cgi-bin/man-cgi?dd+1+NetBSD-5.0.1+i386) . This will   
 erase the MBR (i386) or Sun disk label (sparc64), as well as the NetBSD disk   
 label. If you make a mistake at any point during the RAID setup process, you can   
 always refer to this process to restore the disk to an empty state.  
   
 *Note*: On sparc64, use `/dev/rwd1c` instead of `/dev/rwd1d`!  
   
     # dd if=/dev/zero of=/dev/rwd1d bs=8k count=1  
     1+0 records in  
     1+0 records out  
     8192 bytes transferred in 0.003 secs (2730666 bytes/sec)  
   
 Once this is complete, on i386, verify that both the MBR and NetBSD disk labels   
 are gone. On sparc64, verify that the Sun Disk label is gone as well.  
   
 On i386:  
   
     # fdisk /dev/rwd1d  
       
     fdisk: primary partition table invalid, no magic in sector 0  
     Disk: /dev/rwd1d  
     NetBSD disklabel disk geometry:  
     cylinders: 19386, heads: 16, sectors/track: 63 (1008 sectors/cylinder)  
     total sectors: 19541088  
       
     BIOS disk geometry:  
     cylinders: 1023, heads: 255, sectors/track: 63 (16065 sectors/cylinder)  
     total sectors: 19541088  
       
     Partition table:  
     0: <UNUSED>  
     1: <UNUSED>  
     2: <UNUSED>  
     3: <UNUSED>  
     Bootselector disabled.  
       
     # disklabel -r wd1  
       
     [...snip...]  
     16 partitions:  
     #        size    offset     fstype [fsize bsize cpg/sgs]  
      c:  19541025        63     unused      0     0        # (Cyl.      0* - 19385)  
      d:  19541088         0     unused      0     0        # (Cyl.      0 - 19385)  
   
 On sparc64:  
   
     # sunlabel /dev/rwd1c  
       
     sunlabel: bogus label on `/dev/wd1c' (bad magic number)  
       
     # disklabel -r wd1  
       
     [...snip...]  
     3 partitions:  
     #        size    offset     fstype [fsize bsize cpg/sgs]  
      c:  19541088         0     unused      0     0        # (Cyl.      0 -  19385)  
     disklabel: boot block size 0  
     disklabel: super block size 0  
   
 Now that you are certain the second disk is empty, on i386 you must establish   
 the MBR on the second disk using the values obtained from Disk0/wd0 above. We   
 must remember to mark the NetBSD partition active or the system will not boot.   
 You must also create a NetBSD disklabel on Disk1/wd1 that will enable a RAID   
 volume to exist upon it. On sparc64, you will need to simply   
 [disklabel(8)](http://netbsd.gw.com/cgi-bin/man-cgi?disklabel+8+NetBSD-5.0.1+i386)   
 the second disk which will write the proper Sun Disk Label.  
   
 *Tip*:   
 [disklabel(8)](http://netbsd.gw.com/cgi-bin/man-cgi?disklabel+8+NetBSD-5.0.1+i386)   
 will use your shell' s environment variable `$EDITOR` variable to edit the   
 disklabel. The default is   
 [vi(1)](http://netbsd.gw.com/cgi-bin/man-cgi?vi+1+NetBSD-5.0.1+i386)  
   
 On i386:  
   
     # fdisk -0ua /dev/rwd1d  
     fdisk: primary partition table invalid, no magic in sector 0  
     Disk: /dev/rwd1d  
     NetBSD disklabel disk geometry:  
     cylinders: 19386, heads: 16, sectors/track: 63 (1008 sectors/cylinder)  
     total sectors: 19541088  
       
     BIOS disk geometry:  
     cylinders: 1023, heads: 255, sectors/track: 63 (16065 sectors/cylinder)  
     total sectors: 19541088  
       
     Do you want to change our idea of what BIOS thinks? [n]  
       
     Partition 0:  
     <UNUSED>  
     The data for partition 0 is:  
     <UNUSED>  
     sysid: [0..255 default: 169]  
     start: [0..1216cyl default: 63, 0cyl, 0MB]  
     size: [0..1216cyl default: 19541025, 1216cyl, 9542MB]  
     bootmenu: []  
     Do you want to change the active partition? [n] y  
     Choosing 4 will make no partition active.  
     active partition: [0..4 default: 0] 0  
     Are you happy with this choice? [n] y  
       
     We haven't written the MBR back to disk yet.  This is your last chance.  
     Partition table:  
     0: NetBSD (sysid 169)  
         start 63, size 19541025 (9542 MB, Cyls 0-1216/96/1), Active  
     1: <UNUSED>  
     2: <UNUSED>  
     3: <UNUSED>  
     Bootselector disabled.  
     Should we write new partition table? [n] y  
       
     # disklabel -r -e -I wd1  
     type: unknown  
     disk: Disk1  
     label:  
     flags:  
     bytes/sector: 512  
     sectors/track: 63  
     tracks/cylinder: 16  
     sectors/cylinder: 1008  
     cylinders: 19386  
     total sectors: 19541088  
     [...snip...]  
     16 partitions:  
     #        size    offset     fstype [fsize bsize cpg/sgs]  
      a:  19541025        63       RAID                     # (Cyl.      0*-19385)  
      c:  19541025        63     unused      0     0        # (Cyl.      0*-19385)  
      d:  19541088         0     unused      0     0        # (Cyl.      0 -19385)  
   
 On sparc64:  
   
     # disklabel -r -e -I wd1  
     type: unknown  
     disk: Disk1  
     label:  
     flags:  
     bytes/sector: 512  
     sectors/track: 63  
     tracks/cylinder: 16  
     sectors/cylinder: 1008  
     cylinders: 19386  
     total sectors: 19541088  
     [...snip...]  
     3 partitions:  
     #        size    offset     fstype [fsize bsize cpg/sgs]  
      a:  19541088         0       RAID                     # (Cyl.      0 -  19385)  
      c:  19541088         0     unused      0     0        # (Cyl.      0 -  19385)  
       
     # sunlabel /dev/rwd1c   
     sunlabel> P  
     a: start cyl =      0, size = 19541088 (19386/0/0 - 9541.55Mb)  
     c: start cyl =      0, size = 19541088 (19386/0/0 - 9541.55Mb)  
   
 *Note*: On i386, the `c:` and `d:` slices are reserved. `c:` represents the   
 NetBSD portion of the disk. `d:` represents the entire disk. Because we want to   
 allocate the entire NetBSD MBR partition to RAID, and because `a:` resides   
 within the bounds of `c:`, the `a:` and `c:` slices have same size and offset   
 values and sizes. The offset must start at a track boundary (an increment of   
 sectors matching the sectors/track value in the disk label). On sparc64 however,   
 `c:` represents the entire NetBSD partition in the Sun disk label and `d:` is   
 not reserved. Also note that sparc64's `c:` and `a:` require no offset from the   
 beginning of the disk, however if they should need to be, the offset must start   
 at a cylinder boundary (an increment of sectors matching the sectors/cylinder   
 value).  
   
 ### Initializing the RAID Device  
   
 Next we create the configuration file for the RAID set / volume. Traditionally,   
 RAIDframe configuration files belong in `/etc` and would be read and initialized   
 at boot time, however, because we are creating a bootable RAID volume, the   
 configuration data will actually be written into the RAID volume using the   
 *auto-configure* feature. Therefore, files are needed only during the initial   
 setup and should not reside in `/etc`.  
   
     # vi /var/tmp/raid0.conf  
     START array  
     1 2 0  
       
     START disks  
     absent  
     /dev/wd1a  
       
     START layout  
     128 1 1 1  
       
     START queue  
     fifo 100  
   
 Note that `absent` means a non-existing disk. This will allow us to establish   
 the RAID volume with a bogus component that we will substitute for Disk0/wd0 at   
 a later time.  
   
 Next we configure the RAID device and initialize the serial number to something   
 unique. In this example we use a "YYYYMMDD*`Revision`*" scheme. The format you   
 choose is entirely at your discretion, however the scheme you choose should   
 ensure that no two RAID sets use the same serial number at the same time.  
   
 After that we initialize the RAID set for the first time, safely ignoring the   
 errors regarding the bogus component.  
   
     # raidctl -v -C /var/tmp/raid0.conf raid0  
     Ignoring missing component at column 0  
     raid0: Component absent being configured at col: 0  
              Column: 0 Num Columns: 0  
              Version: 0 Serial Number: 0 Mod Counter: 0  
              Clean: No Status: 0  
     Number of columns do not match for: absent  
     absent is not clean!  
     raid0: Component /dev/wd1a being configured at col: 1  
              Column: 0 Num Columns: 0  
              Version: 0 Serial Number: 0 Mod Counter: 0  
              Clean: No Status: 0  
     Column out of alignment for: /dev/wd1a  
     Number of columns do not match for: /dev/wd1a  
     /dev/wd1a is not clean!  
     raid0: There were fatal errors  
     raid0: Fatal errors being ignored.  
     raid0: RAID Level 1  
     raid0: Components: component0[**FAILED**] /dev/wd1a  
     raid0: Total Sectors: 19540864 (9541 MB)  
     # raidctl -v -I 2009122601 raid0  
     # raidctl -v -i raid0  
     Initiating re-write of parity  
     raid0: Error re-writing parity!  
     Parity Re-write status:  
       
     # tail -1 /var/log/messages  
     Dec 26 00:00:30  /netbsd: raid0: Error re-writing parity!  
     # raidctl -v -s raid0  
     Components:  
               component0: failed  
                /dev/wd1a: optimal  
     No spares.  
     component0 status is: failed.  Skipping label.  
     Component label for /dev/wd1a:  
        Row: 0, Column: 1, Num Rows: 1, Num Columns: 2  
        Version: 2, Serial Number: 2009122601, Mod Counter: 7  
        Clean: No, Status: 0  
        sectPerSU: 128, SUsPerPU: 1, SUsPerRU: 1  
        Queue size: 100, blocksize: 512, numBlocks: 19540864  
        RAID Level: 1  
        Autoconfig: No  
        Root partition: No  
        Last configured as: raid0  
     Parity status: DIRTY  
     Reconstruction is 100% complete.  
     Parity Re-write is 100% complete.  
     Copyback is 100% complete.  
   
 ### Setting up Filesystems  
   
 *Caution*: The root filesystem must begin at sector 0 of the RAID device. If   
 not, the primary boot loader will be unable to find the secondary boot loader.  
   
 The RAID device is now configured and available. The RAID device is a pseudo   
 disk-device. It will be created with a default disk label. You must now   
 determine the proper sizes for disklabel slices for your production environment.   
 For purposes of simplification in this example, our system will have 8.5   
 gigabytes dedicated to `/` as `/dev/raid0a` and the rest allocated to `swap`   
 as `/dev/raid0b`.  
   
 *Caution*: This is an unrealistic disk layout for a production server; the   
 NetBSD Guide can expand on proper partitioning technique. See [Installing   
 NetBSD: Preliminary considerations and preparations*](inst).  
   
 *Note*: Note that 1 GB is 2\*1024\*1024=2097152 blocks (1 block is 512 bytes, or   
 0.5 kilobytes). Despite what the underlying hardware composing a RAID set is,   
 the RAID pseudo disk will always have 512 bytes/sector.  
   
 *Note*: In our example, the space allocated to the underlying `a:` slice   
 composing the RAID set differed between i386 and sparc64, therefore the total   
 sectors of the RAID volumes differs:  
   
 On i386:  
   
      # disklabel -r -e -I raid0  
     type: RAID  
     disk: raid  
     label: fictitious  
     flags:  
     bytes/sector: 512  
     sectors/track: 128  
     tracks/cylinder: 8  
     sectors/cylinder: 1024  
     cylinders: 19082  
     total sectors: 19540864  
     rpm: 3600  
     interleave: 1  
     trackskew: 0  
     cylinderskew: 0  
     headswitch: 0 # microseconds  
     track-to-track seek: 0 # microseconds  
     drivedata: 0  
       
     #        size    offset     fstype [fsize bsize cpg/sgs]  
      a:  19015680         0     4.2BSD      0     0     0  # (Cyl.      0 - 18569)  
      b:    525184  19015680       swap                     # (Cyl.  18570 - 19082*)  
      d:  19540864         0     unused      0     0        # (Cyl.      0 - 19082*)  
   
 On sparc64:  
   
     # disklabel -r -e -I raid0  
     [...snip...]  
     total sectors: 19539968  
     [...snip...]  
     3 partitions:  
     #        size    offset     fstype [fsize bsize cpg/sgs]  
      a:  19251200         0     4.2BSD      0     0     0  # (Cyl.      0 -  18799)  
      b:    288768  19251200       swap                     # (Cyl.  18800 -  19081)  
      c:  19539968         0     unused      0     0        # (Cyl.      0 -  19081)  
   
 Next, format the newly created `/` partition as a 4.2BSD FFSv1 File System:  
   
     # newfs -O 1 /dev/rraid0a  
     /dev/rraid0a: 9285.0MB (19015680 sectors) block size 16384, fragment size 2048  
             using 51 cylinder groups of 182.06MB, 11652 blks, 23040 inodes.  
     super-block backups (for fsck -b #) at:  
     32, 372896, 745760, 1118624, 1491488, 1864352, 2237216, 2610080, 2982944,  
     ...............................................................................  
       
     # fsck -fy /dev/rraid0a  
     ** /dev/rraid0a  
     ** File system is already clean  
     ** Last Mounted on  
     ** Phase 1 - Check Blocks and Sizes  
     ** Phase 2 - Check Pathnames  
     ** Phase 3 - Check Connectivity  
     ** Phase 4 - Check Reference Counts  
     ** Phase 5 - Check Cyl groups  
     1 files, 1 used, 4679654 free (14 frags, 584955 blocks, 0.0% fragmentation)  
   
 ### Migrating System to RAID  
   
 The new RAID filesystems are now ready for use. We mount them under `/mnt` and   
 copy all files from the old system. This can be done using   
 [dump(8)](http://netbsd.gw.com/cgi-bin/man-cgi?dump+8+NetBSD-5.0.1+i386) or   
 [pax(1)](http://netbsd.gw.com/cgi-bin/man-cgi?pax+1+NetBSD-5.0.1+i386).  
   
     # mount /dev/raid0a /mnt  
     # df -h /mnt  
     Filesystem        Size       Used      Avail %Cap Mounted on  
     /dev/raid0a       8.9G       2.0K       8.5G   0% /mnt  
     # cd /; pax -v -X -rw -pe . /mnt  
     [...snip...]  
   
 The NetBSD install now exists on the RAID filesystem. We need to fix the   
 mount-points in the new copy of `/etc/fstab` or the system will not come up   
 correctly. Replace instances of `wd0` with `raid0`.  
   
 The swap should be unconfigured upon shutdown to avoid parity errors on the RAID   
 device. This can be done with a simple, one-line setting in `/etc/rc.conf`.  
   
     # vi /mnt/etc/rc.conf  
     swapoff=YES  
   
 Next, the boot loader must be installed on Disk1/wd1. Failure to install the   
 loader on Disk1/wd1 will render the system un-bootable if Disk0/wd0 fails. You  
 should hope your system won't have to reboot when wd0 fails.  
   
 *Tip*: Because the BIOS/CMOS menus in many i386 based systems are misleading   
 with regard to device boot order. I highly recommend utilizing the `-o   
 timeout=X` option supported by the i386 1st stage boot loader. Setup unique   
 values for each disk as a point of reference so that you can easily determine   
 from which disk the system is booting.  
   
 *Caution*: Although it may seem logical to install the 1st stage boot block into   
 `/dev/rwd1{c,d}` (which is historically correct with NetBSD 1.6.x   
 [installboot(8)](http://netbsd.gw.com/cgi-bin/man-cgi?installboot+8+NetBSD-5.0.1+i386)   
 , this is no longer the case. If you make this mistake, the boot sector will   
 become irrecoverably damaged and you will need to start the process over again.  
   
 On i386, install the boot loader into `/dev/rwd1a`:  
   
     # /usr/sbin/installboot -o timeout=30 -v /dev/rwd1a /usr/mdec/bootxx_ffsv1  
     File system:         /dev/rwd1a  
     Primary bootstrap:   /usr/mdec/bootxx_ffsv1  
     Ignoring PBR with invalid magic in sector 0 of `/dev/rwd1a'  
     Boot options:        timeout 30, flags 0, speed 9600, ioaddr 0, console pc  
   
 On sparc64, install the boot loader into `/dev/rwd1a` as well, however the `-o`   
 flag is unsupported (and un-needed thanks to OpenBoot):  
   
     # /usr/sbin/installboot -v /dev/rwd1a /usr/mdec/bootblk  
     File system:         /dev/rwd1a  
     Primary bootstrap:   /usr/mdec/bootblk  
     Bootstrap start sector: 1  
     Bootstrap byte count:   5140  
     Writing bootstrap  
   
 Finally the RAID set must be made auto-configurable and the system should be   
 rebooted. After the reboot everything is mounted from the RAID devices.  
   
     # raidctl -v -A root raid0  
     raid0: Autoconfigure: Yes  
     raid0: Root: Yes  
     # tail -2 /var/log/messages  
     raid0: New autoconfig value is: 1  
     raid0: New rootpartition value is: 1  
     # raidctl -v -s raid0  
     [...snip...]  
        Autoconfig: Yes  
        Root partition: Yes  
        Last configured as: raid0  
     [...snip...]  
     # shutdown -r now  
   
 ### Warning  
   
 Always use   
 [shutdown(8)](http://netbsd.gw.com/cgi-bin/man-cgi?shutdown+8+NetBSD-5.0.1+i386)   
 when shutting down. Never simply use   
 [reboot(8)](http://netbsd.gw.com/cgi-bin/man-cgi?reboot+8+NetBSD-5.0.1+i386).   
 [reboot(8)](http://netbsd.gw.com/cgi-bin/man-cgi?reboot+8+NetBSD-5.0.1+i386)   
 will not properly run shutdown RC scripts and will not safely disable swap. This   
 will cause dirty parity at every reboot.  
   
 ### The first boot with RAID  
   
 At this point, temporarily configure your system to boot Disk1/wd1. See notes in   
 [[Testing Boot Blocks|guide/rf#adding-text-boot]] for details on this process.   
 The system should boot now and all filesystems should be on the RAID devices.   
 The RAID will be functional with a single component, however the set is not   
 fully functional because the bogus drive (wd9) has failed.  
   
     # egrep -i "raid|root" /var/run/dmesg.boot  
     raid0: RAID Level 1  
     raid0: Components: component0[**FAILED**] /dev/wd1a  
     raid0: Total Sectors: 19540864 (9541 MB)  
     boot device: raid0  
     root on raid0a dumps on raid0b  
     root file system type: ffs  
       
     # df -h  
     Filesystem    Size     Used     Avail Capacity  Mounted on  
     /dev/raid0a   8.9G     196M      8.3G     2%    /  
     kernfs        1.0K     1.0K        0B   100%    /kern  
       
     # swapctl -l  
     Device      1K-blocks     Used    Avail Capacity  Priority  
     /dev/raid0b    262592        0   262592     0%    0  
     # raidctl -s raid0  
     Components:  
               component0: failed  
                /dev/wd1a: optimal  
     No spares.  
     component0 status is: failed.  Skipping label.  
     Component label for /dev/wd1a:  
        Row: 0, Column: 1, Num Rows: 1, Num Columns: 2  
        Version: 2, Serial Number: 2009122601, Mod Counter: 65  
        Clean: No, Status: 0  
        sectPerSU: 128, SUsPerPU: 1, SUsPerRU: 1  
        Queue size: 100, blocksize: 512, numBlocks: 19540864  
        RAID Level: 1  
        Autoconfig: Yes  
        Root partition: Yes  
        Last configured as: raid0  
     Parity status: DIRTY  
     Reconstruction is 100% complete.  
     Parity Re-write is 100% complete.  
     Copyback is 100% complete.  
   
 ### Adding Disk0/wd0 to RAID  
   
 We will now add Disk0/wd0 as a component of the RAID. This will destroy the   
 original file system structure. On i386, the MBR disklabel will be unaffected   
 (remember we copied wd0's label to wd1 anyway) , therefore there is no need to   
 "zero" Disk0/wd0. However, we need to relabel Disk0/wd0 to have an identical   
 NetBSD disklabel layout as Disk1/wd1. Then we add Disk0/wd0 as "hot-spare" to   
 the RAID set and initiate the parity reconstruction for all RAID devices,   
 effectively bringing Disk0/wd0 into the RAID-1 set and "syncing up" both disks.  
   
     # disklabel -r wd1 > /tmp/disklabel.wd1  
     # disklabel -R -r wd0 /tmp/disklabel.wd1  
   
 As a last-minute sanity check, you might want to use   
 [diff(1)](http://netbsd.gw.com/cgi-bin/man-cgi?diff+1+NetBSD-5.0.1+i386) to   
 ensure that the disklabels of Disk0/wd0 match Disk1/wd1. You should also backup   
 these files for reference in the event of an emergency.  
   
     # disklabel -r wd0 > /tmp/disklabel.wd0  
     # disklabel -r wd1 > /tmp/disklabel.wd1  
     # diff /tmp/disklabel.wd0 /tmp/disklabel.wd1  
     # fdisk /dev/rwd0 > /tmp/fdisk.wd0  
     # fdisk /dev/rwd1 > /tmp/fdisk.wd1  
     # diff /tmp/fdisk.wd0 /tmp/fdisk.wd1  
     # mkdir /root/RFbackup  
     # cp -p /tmp/{disklabel,fdisk}* /root/RFbackup  
   
 Once you are sure, add Disk0/wd0 as a spare component, and start reconstruction:  
   
     # raidctl -v -a /dev/wd0a raid0  
     /netbsd: Warning: truncating spare disk /dev/wd0a to 241254528 blocks  
     # raidctl -v -s raid0  
     Components:  
               component0: failed  
                /dev/wd1a: optimal  
     Spares:  
                /dev/wd0a: spare  
     [...snip...]  
     # raidctl -F component0 raid0  
     RECON: initiating reconstruction on col 0 -> spare at col 2  
      11% |****                                   | ETA:    04:26 \  
   
 Depending on the speed of your hardware, the reconstruction time will vary. You   
 may wish to watch it on another terminal (note that you can interrupt  
 `raidctl -S` any time without stopping the synchronisation):  
   
     # raidctl -S raid0  
     Reconstruction is 0% complete.  
     Parity Re-write is 100% complete.  
     Copyback is 100% complete.  
     Reconstruction status:  
       17% |******                                 | ETA: 03:08 -  
   
 After reconstruction, both disks should be *optimal*.  
   
     # tail -f /var/log/messages  
     raid0: Reconstruction of disk at col 0 completed  
     raid0: Recon time was 1290.625033 seconds, accumulated XOR time was 0 us (0.000000)  
     raid0:  (start time 1093407069 sec 145393 usec, end time 1093408359 sec 770426 usec)  
     raid0: Total head-sep stall count was 0  
     raid0: 305318 recon event waits, 1 recon delays  
     raid0: 1093407069060000 max exec ticks  
       
     # raidctl -v -s raid0  
     Components:  
                component0: spared  
                /dev/wd1a: optimal  
     Spares:  
          /dev/wd0a: used_spare  
          [...snip...]  
   
 When the reconstruction is finished we need to install the boot loader on the   
 Disk0/wd0. On i386, install the boot loader into `/dev/rwd0a`:  
   
     # /usr/sbin/installboot -o timeout=15 -v /dev/rwd0a /usr/mdec/bootxx_ffsv1  
     File system:         /dev/rwd0a  
     Primary bootstrap:   /usr/mdec/bootxx_ffsv1  
     Boot options:        timeout 15, flags 0, speed 9600, ioaddr 0, console pc  
   
 On sparc64:  
   
     # /usr/sbin/installboot -v /dev/rwd0a /usr/mdec/bootblk  
     File system:         /dev/rwd0a  
     Primary bootstrap:   /usr/mdec/bootblk  
     Bootstrap start sector: 1  
     Bootstrap byte count:   5140  
     Writing bootstrap  
   
 And finally, reboot the machine one last time before proceeding. This is   
 required to migrate Disk0/wd0 from status "used\_spare" as "Component0" to state   
 "optimal". Refer to notes in the next section regarding verification of clean   
 parity after each reboot.  
   
     # shutdown -r now  
   
 ### Testing Boot Blocks  
   
 At this point, you need to ensure that your system's hardware can properly boot   
 using the boot blocks on either disk. On i386, this is a hardware-dependent   
 process that may be done via your motherboard CMOS/BIOS menu or your controller   
 card's configuration menu.  
   
 On i386, use the menu system on your machine to set the boot device order /   
 priority to Disk1/wd1 before Disk0/wd0. The examples here depict a generic Award   
 BIOS.  
   
 ![Award BIOS i386 Boot Disk1/wd1](/guide/images/raidframe_awardbios2.png)  
 **Award BIOS i386 Boot Disk1/wd1**  
   
 Save changes and exit:  
   
     >> NetBSD/i386 BIOS Boot, Revision 5.2 (from NetBSD 5.0.2)  
     >> (builds@b7, Sun Feb 7 00:30:50 UTC 2010)  
     >> Memory: 639/130048 k  
     Press return to boot now, any other key for boot menu  
     booting hd0a:netbsd - starting in 30  
   
 You can determine that the BIOS is reading Disk1/wd1 because the timeout of the   
 boot loader is 30 seconds instead of 15. After the reboot, re-enter the BIOS and   
 configure the drive boot order back to the default:  
   
 ![Award BIOS i386 Boot Disk0/wd0](/guide/images/raidframe_awardbios1.png)  
 **Award BIOS i386 Boot Disk0/wd0**  
   
 Save changes and exit:  
   
     >> NetBSD/i386 BIOS Boot, Revision 5.2 (from NetBSD 5.0.2)  
     >> Memory: 639/130048 k  
     Press return to boot now, any other key for boot menu  
     booting hd0a:netbsd - starting in 15  
   
 Notice how your custom kernel detects controller/bus/drive assignments   
 independent of what the BIOS assigns as the boot disk. This is the expected   
 behavior.  
   
 On sparc64, use the Sun OpenBoot **devalias** to confirm that both disks are bootable:  
   
     Sun Ultra 5/10 UPA/PCI (UltraSPARC-IIi 400MHz), No Keyboard  
     OpenBoot 3.15, 128 MB memory installed, Serial #nnnnnnnn.  
     Ethernet address 8:0:20:a5:d1:3b, Host ID: nnnnnnnn.  
       
     ok devalias  
     [...snip...]  
     cdrom /pci@1f,0/pci@1,1/ide@3/cdrom@2,0:f  
     disk /pci@1f,0/pci@1,1/ide@3/disk@0,0  
     disk3 /pci@1f,0/pci@1,1/ide@3/disk@3,0  
     disk2 /pci@1f,0/pci@1,1/ide@3/disk@2,0  
     disk1 /pci@1f,0/pci@1,1/ide@3/disk@1,0  
     disk0 /pci@1f,0/pci@1,1/ide@3/disk@0,0  
     [...snip...]  
       
     ok boot disk0 netbsd  
     Initializing Memory [...]  
     Boot device /pci/pci/ide@3/disk@0,0 File and args: netbsd  
     NetBSD IEEE 1275 Bootblock  
     >> NetBSD/sparc64 OpenFirmware Boot, Revision 1.13  
     >> (builds@b7.netbsd.org, Wed Jul 29 23:43:42 UTC 2009)  
     loadfile: reading header  
     elf64_exec: Booting [...]  
     symbols @ [....]  
      Copyright (c) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005,  
          2006, 2007, 2008, 2009  
          The NetBSD Foundation, Inc.  All rights reserved.  
      Copyright (c) 1982, 1986, 1989, 1991, 1993  
          The Regents of the University of California.  All rights reserved.  
     [...snip...]  
   
 And the second disk:  
   
     ok boot disk2 netbsd  
     Initializing Memory [...]  
     Boot device /pci/pci/ide@3/disk@2,0: File and args:netbsd  
     NetBSD IEEE 1275 Bootblock  
     >> NetBSD/sparc64 OpenFirmware Boot, Revision 1.13  
     >> (builds@b7.netbsd.org, Wed Jul 29 23:43:42 UTC 2009)  
     loadfile: reading header  
     elf64_exec: Booting [...]  
     symbols @ [....]  
      Copyright (c) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005,  
          2006, 2007, 2008, 2009  
          The NetBSD Foundation, Inc.  All rights reserved.  
      Copyright (c) 1982, 1986, 1989, 1991, 1993  
          The Regents of the University of California.  All rights reserved.  
     [...snip...]  
   
 At each boot, the following should appear in the NetBSD kernel   
 [dmesg(8)](http://netbsd.gw.com/cgi-bin/man-cgi?dmesg+8+NetBSD-5.0.1+i386) :  
   
     Kernelized RAIDframe activated  
     raid0: RAID Level 1  
     raid0: Components: /dev/wd0a /dev/wd1a  
     raid0: Total Sectors: 19540864 (9541 MB)  
     boot device: raid0  
     root on raid0a dumps on raid0b  
     root file system type: ffs  
   
 Once you are certain that both disks are bootable, verify the RAID parity is   
 clean after each reboot:  
   
     # raidctl -v -s raid0  
     Components:  
               /dev/wd0a: optimal  
               /dev/wd1a: optimal  
     No spares.  
     [...snip...]  
     Component label for /dev/wd0a:  
        Row: 0, Column: 0, Num Rows: 1, Num Columns: 2  
        Version: 2, Serial Number: 2009122601, Mod Counter: 67  
        Clean: No, Status: 0  
        sectPerSU: 128, SUsPerPU: 1, SUsPerRU: 1  
        Queue size: 100, blocksize: 512, numBlocks: 19540864  
        RAID Level: 1  
        Autoconfig: Yes  
        Root partition: Yes  
        Last configured as: raid0  
     Component label for /dev/wd1a:  
        Row: 0, Column: 1, Num Rows: 1, Num Columns: 2  
        Version: 2, Serial Number: 2009122601, Mod Counter: 67  
        Clean: No, Status: 0  
        sectPerSU: 128, SUsPerPU: 1, SUsPerRU: 1  
        Queue size: 100, blocksize: 512, numBlocks: 19540864  
        RAID Level: 1  
        Autoconfig: Yes  
        Root partition: Yes  
        Last configured as: raid0  
     Parity status: clean  
     Reconstruction is 100% complete.  
     Parity Re-write is 100% complete.  
     Copyback is 100% complete.  
   

Removed from v.1.4  
changed lines
  Added in v.1.14


CVSweb for NetBSD wikisrc <wikimaster@NetBSD.org> software: FreeBSD-CVSweb