Assumptions:
1. Global zone name: global
2. Non-global zone name: ngzname
3. Disk you want mounted: c1t1d1s0
4. Non-global zone is up and running

Let’s get started.

1. Identify the disk you want to mount from the global to the zone. In this case its c1t1d1s0.

2. Mount the disk as you normally would from the global zone.
a. Create the mount point from the global zone

        # mkdir /mnt/olddisk

b. Create the mount point for the Non-global zone

        # mkdir /zones/ngzname/root/mnt/olddisk

3. Mount the disk to both the global and Non-global zone
a. For the global

        # mount /dev/dsk/c1t1d1s0 /mnt/olddisk

b. For the Non-global
ENSURE ngzone is up and running. For me, mounting the file system as read-only was a requirement. Your mileage may vary.

        # mount -o ro -F lofs /mnt/olddisk /zones/ngzname/root/mnt/olddisk

4. Verify mount point is up and running
a. From global

         # df -h /mnt/olddisk
         # mount | grep /mnt/olddisk

b. Log into Non-global zone

          # df -h /mnt/olddisk

The disk should be mount as read only from the global to the Non-global zone as read-only. Enjoy!

Assume the following is true:

1. Display file system mount status

# df -h /mnt/vol1/
Filesystem             size   used  avail capacity  Mounted on
/dev/zvol/dsk/fort/vol1
                       962M   1.0M   903M     1%    /mnt/vol1

Notice its 1Gig in size (962M formatted).

2. Display current ZFS volume size and reservation properties

# zfs get volsize,reservation fort/vol1
NAME           PROPERTY     VALUE          SOURCE
fort/vol1  volsize      1G             -
fort/vol1  reservation  1G             local

Confirms the volume is 1Gig in size with a 1Gig reservation.

3. Change the volsize ZFS property for fort/vol1 to 2Gig.

# zfs set volsize=2g fort/vol1

4. Confirm the changes were made by displaying the volsize and reservation ZFS properties.

# zfs get volsize,reservation fort/vol1
NAME           PROPERTY     VALUE          SOURCE
fort/vol1  volsize      2G             -
fort/vol1  reservation  2G             local

Notice the “volsize” is now set to 2Gig. At this point, the OS doesnt see the expanded file system. In order for the OS to see the size, you will have to expand the file system with “growfs”.

5. Run growfs to expand mounted file system

# growfs -M /mnt/vol1 /dev/zvol/rdsk/fort/vol1
Warning: 2048 sector(s) in last cylinder unallocated
/dev/zvol/rdsk/fort/vol1:   4194304 sectors in 683 cylinders of 48 tracks, 128 sectors
        2048.0MB in 49 cyl groups (14 c/g, 42.00MB/g, 20160 i/g)
super-block backups (for fsck -F ufs -o b=#) at:
 32, 86176, 172320, 258464, 344608, 430752, 516896, 603040, 689184, 775328,
 3359648, 3445792, 3531936, 3618080, 3704224, 3790368, 3876512, 3962656,
 4048800, 4134944

The output should now show you the new volume size.

6. Confirm the size of mounted file system using df or dd.

# df -h /mnt/vol1/
Filesystem             size   used  avail capacity  Mounted on
/dev/zvol/dsk/fort/vol1
                       1.9G   2.0M   1.8G     1%    /mnt/vol1

# dd if=/dev/zvol/dsk/fort/vol1 of=/dev/null bs=1024k
2048+0 records in
2048+0 records out

Thats it! Growing a ZFS volume is an online operation in my mind.

The method below describes a way in which you can mount a new file system (UFS) to a running Solaris 10 zone without the need of a reboot.

Assumptions:
• All commands need to be run from the global zone as the root user unless otherwise specified.

Prerequisites:

• A Solaris 10 host with at least one running non-global zone.

Limitations:

• You can not unmount the file system from the non-global zone. Some say this is a benefit

Lets get started!

Log into the global zone and create the new file system like you normally would. In this example, I use SVM to create a new metadevice (d1000), one way mirror (d2000), and a 1Gig soft partition (d2001). Your mileage may vary.


# metainit d1000 1 1 c1t5d0s0 -i 128k
# metainit d2000 -m d1000
# metainit d2001 -p d2000 1g
# echo y | newfs -i 8192 /dev/md/rdsk/d2001


Create the block and character devices for the zone. Run “ls -lL” against the newly created soft partition (d2001) to find the major and minor character and block values.


# ls -lL /dev/md/*dsk/d2001
brw-r----- 1 root sys 85, 2001 Oct 6 13:46 /dev/md/dsk/d2001
crw-r----- 1 root sys 85, 2001 Oct 6 13:47 /dev/md/rdsk/d2001


NOTE: The major and minor values are 85 and 2001, respectively. Notice that the “dsk” device is a “block-type” file and “rdsk” device is a “character-type” file.

Run mknod to create the two special files for the zone. Keep in mind that the name of the two files need to be saved down the myzone device path.


# mknod /zones/myzone1/dev/md/dsk/d2001 b 85 2001
# mknod /zones/myzone1/dev/md/rdsk/d2001 c 85 2001


Correct the permissions and ownerships of the new d2001 block-type and character-type files.


# chmod 640 /zones/myzone1/dev/md/*dsk/d2001
# chown root:root /zones/myzone1/dev/md/*dsk/d2001


Create a mount point in the “myzone1″ non-global zone.


# mkdir /zones/myzone1/root/wandisimo


Confirm its not already mounted


# mount | grep wandisimo


You should not get anything back from the command above.

Mount the file system


# mount /dev/md/dsk/d2001 /zones/myzone1/root/wandisimo


You are probably thinking to your self, “Why couldn’t we just skip to this step and be done with it?” From my testing, if the block and character files are not created down the “myzone1″ device path first, it fails to mount. If you were to add the soft partition using zonecfg or by adding it to the zone XML file then rebooting the non-global zone, the boot process actually creates the block and character device for you.

NOTE: Even though the file system successfully gets mounted, it will not show up in a “df” output as actually being mounted.

Check to see if its mounted


# df -h | grep wandisimo


You should not get any output.

Confirm the file system is in fact mounted by running the mount command


# mount | grep wandisimo
/zones/myzone1/root/wandisimo on /dev/md/dsk/d2001 read/write/setuid/devices/intr/largefiles/logging/xattr/onerror=panic/dev=15407d3 on Tue Oct 6 14:11:45 2009


Log into the running non-global zone (myzone1) to confirm the file system is indeed mounted


# zlogin myzone1
(myzone1)# df -h /wandisimo
Filesystem size used avail capacity Mounted on
/wandisimo 1008M 1.0M 947M 1% /wandisimo


Done. No need to reboot the non-global zone!

If you would like to make this permanent, you will need to add it to the /etc/vfstab in the global zone.

I started with 500Meg disks and worked my way down to 50Meg disks. Keep in mind that these are not real disks. They are empty files created by the mkfile command and stored in /tmp/ that I am using as disks. I would not recommend using files in a live system but they work great for tests and experiments.

All my tests were as expected until I created the 50Meg disk/files.

Using the Bash shell, here is what I did:

1. Create the 50Meg disk files

# for i in {1..5}; do mkfile 50m /tmp/disk${i}; done

This created 5 x 50Meg files in /tmp.

2. I then attempted to create a raidz ZFS pool using those files as my disks.

# zpool create fort raidz /var/tmp/{1..5}
cannot create 'fort': one or more devices is less than the minimum size (64M)

3. Changing the size to 64Meg files worked.

#  for i in {1..5}; do mkfile 64m /tmp/disk${i}; done
# zpool create fort raidz /tmp/disk{1..5}
# zpool list
NAME                    SIZE    USED   AVAIL    CAP  HEALTH     ALTROOT
fort                    296M    143K    296M     0%  ONLINE     -

# zfs list
NAME   USED  AVAIL  REFER  MOUNTPOINT
fort   109K   204M  39.1K  /fort

# zpool status -v
  pool: fort
 state: ONLINE
 scrub: none requested
config:

        NAME            STATE     READ WRITE CKSUM
        fort            ONLINE       0     0     0
          raidz1        ONLINE       0     0     0
            /tmp/disk1  ONLINE       0     0     0
            /tmp/disk2  ONLINE       0     0     0
            /tmp/disk3  ONLINE       0     0     0
            /tmp/disk4  ONLINE       0     0     0
            /tmp/disk5  ONLINE       0     0     0

errors: No known data errors

For future reference, your disks must be a equal to or greater than 64Megs in size. Great for 64Meg USB drives, sarcasm included…

could not verify fs /test: zfs ‘tank/test’ mountpoint is not “legacy”
zoneadm: zone non-global-zone failed to verify

The fix would be to update/create the ZFS dataset as a legacy mount point. For this post, I assume that you have both basic understanding of Solaris zone technology and zfs file systems.

In the version of Solaris that I run, (Solaris 10 11/06, sparc) here are the steps you would need to follow in order to add a ZFS file system to a Solaris non-global zone.

1. Create a legacy zfs pool from the global zone
   

   # zpool create tank c0t0d0s0

2. Create a zfs dataset using the new tank zfs pool
   

   # zfs create tant/test
   

3. Display the zfs pool
   

   # zfs list
   NAME                USED  AVAIL  REFER  MOUNTPOINT
   tank               6.57G   120G  26.5K  /tank
   tank/test          24.5K   120G  24.5K  /tank/test
   

4. Set the new tank/test dataset as a legacy mount point
   

   # zfs set mountpoint=legacy tank/test
   

5. Display the zfs pool
   

   # zfs list
   NAME                USED  AVAIL  REFER  MOUNTPOINT
   tank               6.57G   120G  24.5K  /tank
   tank/test          24.5K   120G  24.5K  legacy
   

Notice the mount point for tank/test changed from the actual mount point /tank/test to legacy.

6. Create a Solaris non-global zone with similar options as below
   

   # zonecfg -z non-global-zone info
   
   ...
   fs:
           dir: /test
           special: tank/test
           raw not specified
           type: zfs
           options: []
   dataset:
           name: tank/test
   ...
   

After the zone has been created and booted, login into the non-global-zone and running a “df” will show you the new ZFS pool is mounted and ready to satisfy your ZFS needs.

The key line to adding the SVM Metadevice to the non-global zone is the following:

zonecfg:myzone:net> add device
zonecfg:myzone:net> set match=/dev/md/*dsk/d610
zonecfg:myzone:net> end

I used an asterisk to signify that I want both the the dsk and rdsk devices.

Assuming all went well, you should be able to mount and unmount the SVM Metadevice from the non-global zone. Why would you want to provide the ability for a non-global zone to mount and un-mount a file system at will? In my case, the mount point assigned to d110 in the example above is cloned daily using EMC clone technology.

When the box came back up I was pleasantly surprised to see the zfs pool already mounted and online. I didn’t remember whether or not I had added the entry in /etc/vfstab. Scratching my head, I immediately went off to check /etc/vfstab on the box but there was no entry for the zfs pool I was working on. Hmm…

How did the zfs pool automatically mount itself at boot time? Was it magical zfs gnomes? How about the Sun Oracle Gods?

The Solaris 10 OS reads /etc/zfs/zpool.cache at boot time and mounts all the pools it finds in that file. That file is in binary form so if you want to take a peek, run strings against that file.

# /usr/bin/strings /etc/zfs/zpool.cache
tank
version
name
tank
state
pool_guid
(7id
vdev_tree
type
root
guid
(7id
children
type
disk
guid
/§y[ü
path
/dev/dsk/c1t5d0s0
devid
id1,ssd@n20000011c6f0b1fb/a
whole_disk
metaslab_array
metaslab_shift
ashift
asize

If you wanted to get fancy, you can even set a zfs property for the pool so that it doesn’t mount automatically at boot time using the “canmount” option. Further details can be found at the zfs man page http://docs.sun.com/app/docs/doc/819-2240/zfs-1m