Mirroring Disk With Solaris Disksuite (formerly Solstice)

The first step to setting up mirroring using DiskSuite is to install the DiskSuite packages and any necessary patches for systems prior to Solaris 9. SVM is part of the base system in Solaris 9. The latest recommended version of DiskSuite is 4.2 for systems running Solaris 2.6 and Solaris 7, and 4.2.1 for Solaris 8. There are currently three packages and one patch necessary to install DiskSuite 4.2. They are:

SUNWmd (Required)
SUNWmdg (Optional GUI)
SUNWmdn (Optional SNMP log daemon)
106627-19 (obtain latest revision)

The packages should be installed in the same order as listed above. Note that a reboot is necessary after the install as new drivers will be added to the Solaris kernel. For DiskSuite 4.2.1, install the following packages:

SUNWmdu (Commands)
SUNWmdr (Drivers)
SUNWmdx (64-Bit Drivers)
SUNWmdg (Optional GUI)
SUNWmdnr (Optional log daemon configs)
SUNWmdnu (Optional log daemon)

For Solaris 2.6 and 7, to make life easier, be sure to update your PATH and MANPATH variables to add DiskSuite's directories. Executables reside in /usr/opt/SUNWmd/sbin and man pages in /usr/opt/SUNWmd/man. In Solaris 8, DiskSuite files were moved to "normal" system locations (/usr/sbin) so path updates are not necessary.

The Environment
In this example we will be mirroring two disks, both on the same controller. The first disk will be the primary disk and the second will be the mirror. The disks are:

Disk 1: c0t0d0
Disk 2: c0t1d0

The partitions on the disks are presented below. There are a few items of note here. Each disk is partitioned exactly the same. This is necessary to properly implement the mirrors. Slice 2, commonly referred to as the 'backup' slice, which represents the entire disk must not be mirrored. There are situations where slice 2 is used as a normal slice, however, this author would not recommend doing so. The three unassigned partitions on each disk are configured to each be 10MB. These 10MB slices will hold the DiskSuite State Database Replicas, or metadbs. More information on the state database replicas will be presented below. In DiskSuite 4.2 and 4.2.1, a metadb only occupies 1034 blocks (517KB) of space. In SVM, they occupy 8192 blocks (4MB). This can lead to many problems during an upgrade if the slices used for the metadb replicas are not large enough to support the new larger databases.

Disk 1:
c0t0d0s0: /
c0t0d0s1: swap
c0t0d0s2: backup
c0t0d0s3: unassigned
c0t0d0s4: /var
c0t0d0s5: unassigned
c0t0d0s6: unassigned
c0t0d0s7: /export

Disk 2:
c0t1d0s0: /
c0t1d0s1: swap
c0t1d0s2: backup
c0t1d0s3: unassigned
c0t1d0s4: /var
c0t1d0s5: unassigned
c0t1d0s6: unassigned
c0t1d0s7: /export

The Database State Replicas

The database state replicas serve a very important function in DiskSuite. They are the repositories of information on the state and configuration of each metadevice (A logical device created through DiskSuite is known as a metadevice). Having multiple replicas is critical to the proper operation of DiskSuite.

· There must be a minimum of three replicas. DiskSuite requires at least half of the replicas to be present in order to continue to operate.
· 51% of the replicas must be present in order to reboot.
· Replicas should be spread across disks and controllers where possible.
· In a three drive configuration, at least one replica should be on each disk, thus allowing for a one disk failure.
· In a two drive configuration, such as the one we present here, there must be at least two replicas per disk. If there were only three and the disk which held two of them failed, there would not be enough information for DiskSuite to function and the system would panic.

Here we will create our state replicas using the metadb command:

# metadb -a -f /dev/dsk/c0t0d0s3
# metadb -a /dev/dsk/c0t0d0s5
# metadb -a /dev/dsk/c0t0d0s6
# metadb -a /dev/dsk/c0t1d0s3
# metadb -a /dev/dsk/c0t1d0s5
# metadb -a /dev/dsk/c0t1d0s6

The -a and -f options used together create the initial replica. The -a option attaches a new database device and automatically edits the appropriate files.

Initializing Submirrors

Each mirrored meta device contains two or more submirrors. The meta device gets mounted by the operating system rather than the original physical device. Below we will walk through the steps involved in creating metadevices for our primary filesystems. Here we create the two submirrors for the / (root) filesystem, as well as a one way mirror between the meta device and its first submirror.

# metainit -f d10 1 1 c0t0d0s0
# metainit -f d20 1 1 c0t1d0s0
# metainit d0 -m d10

The first two commands create the two submirrors. The -f option forces the creation of the submirror even though the specified slice is a mounted filesystem. The second two options 1 1 specify the number of stripes on the metadevice and the number of slices that make up the stripe. In a mirroring situation, this should always be 1 1. Finally, we specify the logical device that we will be mirroring.

After mirroring the root partition, we need to run the metaroot command. This command will update the root entry in /etc/vfstab with the new metadevice as well as add the appropriate configuration information into /etc/system. Ommitting this step is one of the most common mistakes made by those unfamiliar with DiskSuite. If you do not run the metaroot command before you reboot, you will not be able to boot the system!

# metaroot d0

Next, we continue to create the submirrors and initial one way mirrors for the metadevices which will replace the swap, and /var partitions.

# metainit -f d11 1 1 c0t0d0s1
# metainit -f d21 1 1 c0t1d0s1
# metainit d1 -m d11
# metainit -f d14 1 1 c0t0d0s4
# metainit -f d24 1 1 c0t1d0s4
# metainit d4 -m d14
# metainit -f d17 1 1 c0t0d0s7
# metainit -f d27 1 1 c0t1d0s7
# metainit d7 -m d17

Updating /etc/vfstab

The /etc/vfstab file must be updated at this point to reflect the changes made to the system. The / partition will have already been updated through the metaroot command run earlier, but the system needs to know about the new devices for swap and /var. The entries in the file will look something like the following:

/dev/md/dsk/d1 - - swap - no -
/dev/md/dsk/d4 /dev/md/rdsk/d4 /var ufs 1 yes -
/dev/md/dsk/d7 /dev/md/rdsk/d7 /export ufs 1 yes -

Notice that the device paths for the disks have changed from the normal style
/dev/dsk/c#t#d#s# and /dev/rdsk/c#t#d#s# to the new metadevice paths,
/dev/md/dsk/d# and /dev/md/rdsk/d#.

The system can now be rebooted. When it comes back up it will be running off of the new metadevices. Use the df command to verify this. In the next step we will attach the second half of the mirrors and allow the two drives to synchronize.

Attaching the Mirrors
Now we must attach the second half of the mirrors. Once the mirrors are attached it will begin an automatic synchonization process to ensure that both halves of the mirror are identical. The progress of the synchonization can be monitored using the metastat command. To attach the submirrors, issue the following commands:

# metattach d0 d20
# metattach d1 d21
# metattach d4 d24
# metattach d7 d27

Final Thoughts

With an eye towards recovery in case of a future disaster it may be a good idea to find out the physical device path of the root partition on the second disk in order to create an Open Boot PROM (OBP) device alias to ease booting the system if the primary disk fails.

In order to find the physical device path, simply do the following:

# ls -l /dev/dsk/c0t1d0s0

This should return something similar to the following:

/sbus@3,0/SUNW,fas@3,8800000/sd@1,0:a

Using this information, create a device alias using an easy to remember name such as altboot. To create this alias, do the following in the Open Boot PROM:
ok nvalias altboot /sbus@3,0/SUNW,fas@3,8800000/sd@1,0:a

It is now possible to boot off of the secondary device in case of failure using boot altboot from the OBP.