Document Body:

Currently the task of setting up automatic failover on a SPARC(R) is
relatively straight forward, however the same task on an x86 machine is
considerably more convoluted and difficult. For a SPARC machine, only a
couple of simple commands to set the nvalias in the NVRAM are needed to set
up automatic failover to the secondary drive. To set up automatic failover
on an x86 machine, the System Administrator needs to modify two files on
the boot floppy (boot.rc and bootenv.rc) and ALWAYS boot from the floppy.

Setting up automatic failover for a mirrored root:

Currently for automatic failover to succeed this solution requires use of
the boot floppy. The files boot.rc and bootenv.rc on the boot floppy need
to be modified and autoboot must be set. The floppy is used as a primitive
type of boot prom, as Solaris does not have control over the varied types
of BIOS that can be found on these machines, to allow the system to
automatically failover to the alternate boot device.

To set up autofailover use the following procedure:

1) Create the boot floppy from Solaris CD2 using dd.exe.

2) Modify the /floppy/floppy0/solaris/boot.rc file:

  The following line in /floppy/floppy0/boot.rc:


    run /boot/solaris/bootconf.exe ${confflags}


  should be changed to:


    run /boot/solaris/bootconf.exe -n ${confflags}


  Add these lines immediately after the previous line:


    cd /options
    if .streq ( ${root-is-mounted}X , falseX )
      echo
      echo 'Failed to mount primary bootpath:'
      echo "  ${bootpath}"
      echo
      echo 'Trying alternate boot disk:'
      echo "  ${altbootpath}"
      echo
      setprop bootpath "${altbootpath}"
      run /boot/solaris/bootconf.exe ${confflags}
    endif


3) Modify the /floppy/floppy0/solaris/bootenv.rc file as follows.

  Make sure auto-boot? is set to true:


    setprop auto-boot? True


  Set auto-boot-timeout to an appropriately low value:


    setprop auto-boot-timeout 5


  Set altbootpath to your alternate submirror:


    setprop altbootpath <backup_root_device_id>


  Make sure the bootpath is set to the location of your primary root  
  partition:


    setprop bootpath <primary_root_device_id>


  e.g.
    setprop auto-boot? true
    setprop auto-boot-timeout 5
    setprop bootpath    
      /pci@0,0/pci8086,2545@3/pci8086,1460@1d/pci8086,341a@7,1/sd@0,0:a
    setprop altbootpath
      /pci@0,0/pci8086,2545@3/pci8086,1460@1d/pci8086,341a@7,1/sd@1,0:a


SVM Failed Disk Recovery

When one of the disks fails, reboot the system with the boot floppy inserted.
During the boot you will be told that there are state databases which no longer exist.
You will then be prompted to give the root password for system maintenance.

1) Enter the root password for system maintenance.

2) Execute metadb to see which state databases have failed.

3) Delete references to the dead state databases:

    Metadb -d <failed disk> (eg c0t1d0s3)


4) Reboot the system which should come up normally, however you will see
some messages reporting that, for example, metadevice d21 is "unavailable
needs maintenance" for each of the slices which were on the failed disk.

5) Execute metastat and make a note of the slices and their disk slice
locations.

6) Shutdown the server, replace the failed drive with a replacement and power up again.

7) Create a copy of the good disk's partition information onto the new disk:

     fdisk -W /var/tmp/fdisk.dat  /dev/rdsk/c0t0d0p0  <good disk>
     fdisk -F /var/tmp/fdisk.dat  /dev/rdsk/c0t1d0p0  <replaced disk>


8) Add bootblock as per answerbook to the new drive.

  http://docs.sun.com/db/doc/806-4073/...isksxadd-45774


e.g.
installboot /usr/platform/`uname -i`/lib/fs/ufs/pboot
/usr/platform/`uname -i`/lib/fs/ufs/bootblk /dev/rdsk/c0t1d0s2 <replaced disk>

9) Copy the disk layout of the good disk with the new replacement for the failed disk:

 prtvtoc /dev/rdsk/<good disk> | fmthard -s - /dev/rdsk/<replaced disk>


10) Now create new state databases on the new disk:

       metadb -a -c3 <replaced disk>


11) Finally you have to execute metareplace for each failed slice:

       metareplace -e <SVM device> <replaced disk>


   e.g.        
       metareplace -e d30 c0t1d0s0
       metareplace -e d31 c0t1d0s1
       metareplace -e d32 c0t1d0s4
       metareplace -e d33 c0t1d0s5
       metareplace -e d34 c0t1d0s7          


You can monitor the progress of the re-sync by either using the metastat
command or by firing up the Solaris Management Console (SMC) and double
clicking on "The Computer" -> "Storage" -> "Enhanced Storage" -> "Volumes".

Issues:

There is an issue with reference to booting from the second half of an SVM
mirror on certain systems x86 systems. For example, with a failed first
half of a mirror on a V60x the system should boot from the second half of
the mirror, but instead fails with an error like the following:

 unable to mount a Solaris root file system from the device: 
 Disk target:SEGATE ST336607LSUN36G 0307 on Adaptec Ultra/320 
 on Board PCI bus 4, at Dev7, func1 Error message from mount:


 /pci@0,0/pci8086,2545@3/pci8086,1460@1d/pci8086,341a@7,1/sd@1,0:a


Workaround:

The easiest solution is to pull the failed first disk and put the second
disk in its place. The replacement disk can then be put in the free slot
and sync(ed) from there.