3 ways to unencapsulate (including manual unencapsulation)
Article ID: 100005929
Updated On:
Resolution
There are 3 ways of unencapsulating the root disk:
[1.] Unencapsulating the simple way via 'vxunroot'
(/usr/lib/vxvm/bin/vxunroot).
[2.] Unencapsulating manually while Veritas Volume Manager
(VxVM) is up and running by either using the 'vxva'
GUI or 'vxedit' CLI.
[3.] Unencapsulating manually while booted from cdrom.
Note: Options [1.] and [2.] require that you be booted from
either the root disk or the mirror disk and that VxVM
be up and running at that time.
For obvious reasons of requiring VxVM to be up and
running, there may be instances when this is not
feasible and thus option [3.] is the most ideal.
Below is a step-by-step procedure on how to go
about unencapsulating the root disk manually while
booted from cdrom.
=-=-=-=-=-=-=-=-=
STEPS TO FOLLOW:
=-=-=-=-=-=-=-=-=
1. Boot the system from your Solaris CD media in your cdrom drive.
First, shutdown the system to get to the Open Boot PROM level.
# /usr/sbin/shutdown -y -g0 -i0
At The OBP (Open Boot PROM) level issue the following command:
ok boot cdrom -sw
where "-s" means single-user mode and "w",
writeable mode).
2. As an added precaution, if you feel you need to backup the
data off of your boot drive partitions, you may do so only
for the so-called "big-4" partitions which are /, swap, /usr
/var, (obviously since swap is a tmpfs and not a ufs
you do NOT need to back this up) BUT is NOT necessary since
you will essentially be backing up the same data on
Veritas volumes (when you encapsulated the boot drive)...
which means if you will be restoring this data, you will
STILL have to go thru this same process of unencapsulating
it and re-encapsulating your boot drive.
Why only the big-4 partitions you ask and not "data/application"
partitions? Well, this is precisely because of the fact that when
you encapsulate a root disk, the Solaris hard partitions for the
"big-4" partitions ALWAYS get retained even after an encapsulation
process and the other "non-big-4" partitions get removed and Volume
Manager then assigns it's own 2 partitions - the "private" and
"public" regions (usually slices 3 and 4 BUT is really dependent
on the free or available partitions prior to the encapsulation process).
Whereas for non-root disks, Volume Manager retains nothing for the
Solaris partitioning scheme and instead again creates its own
2 partitions - the "private" and "public" regions - but instead
blows away all other partitions. When I say "blows away all other
partitions" I don't necessary mean that data gets lost or moved.
In both instances DATA NEVER GETS MOVED or LOST!
This will be discussed in detail later on.
While still booted from cdrom do the following:
# mt -f /dev/rmt/on stat
# ufsdump 0f /dev/rmt/0n /dev/rdsk/c#t#d#s0
(example slice for /)
# ufsdump 0f /dev/rmt/0n /dev/dsk/c#t#d#s5
(example slice for /var)
# ufsdump 0f /dev/rmt/0n /dev/dsk/c#t#d#s6
(example slice for /usr)
3. Fsck the root filesystem before mounting it.
# fsck /dev/rdsk/c#t#d#s0
where c#t#d#s0 is your root
partition say, c0t0d0s0 as an example.
4. Mount the root filesystem on /a.
# mount /dev/dsk/c#t#d#s0 /a
where c#t#d#s0 is still your root
partition say, c0t0d0s0 as an example.
5. Make copies of your /a/etc/system & /a/etc/vfstab files to ensure
that you have backup copies of them in the event that you need to
reference or go back to these files before the edits you make to them.
# cd /a/etc
# cp system system.ORIG
# cp vfstab vfstab.ORIG
6. Edit your /a/etc/system & /a/etc/vfstab files back to their original
states prior to the encapsulation process.
You will find a /a/etc/vfstab.prevm file. This is your original vfstab
file before Veritas Volume Manager (VxVM) took control of your boot drive.
Use this file to rebuild your current vfstab file so that it reflects
the Solaris hard partitions for the boot slices (/, swap, /usr, /var,
/opt as examples), rather than the Volume Manager volumes.
# cp /a/etc/vfstab.prevm /a/etc/vfstab
Note: Make sure that the vfstab file reflects the old Solaris hard
partitions for the boot disk. Ensure that there are NO
"/dev/vx" devices at this point within this file. If there are
any "/dev/vx" devices, comment them out with a "#".
Edit your /a/etc/system file and remove the 2 entries listed below:
rootdev:/pseudo/vxio@0:0
set vxio:vol_rootdev_is_volume=1
# vi /a/etc/system
* vxvm_START (do not remove)
forceload: drv/vxdmp
forceload: drv/vxio
forceload: drv/vxspec
forceload: drv/sd
forceload: drv/esp
forceload: drv/dma
forceload: drv/sbi
forceload: drv/io-unit
forceload: drv/sf
forceload: drv/pln
forceload: drv/soc
forceload: drv/socal
rootdev:/pseudo/vxio@0:0
set vxio:vol_rootdev_is_volume=1
* vxvm_END (do not remove)
Note: REMOVE by deleting ONLY the following lines:
rootdev:/pseudo/vxio@0:0
set vxio:vol_rootdev_is_volume=1
You MUST remove them and not just comment them
out via an "*". Use the "dd" (delete line) function of
the "vi" editor. Do NOT touch the other entries!
These 2 entries will automatically be put back in
once the root drive is re-encapulated.
7. Since the root drive's big-4 partitions (/, swap, /usr, /var)
ALWAYS gets retained even after an encapsulation process you
will NEED to recover the hard partitions for the following:
7.A. "Non-big-4" partitions that are part of the primary boot disk.
(Example: /opt, /home, /export/home assuming that these
partitions all reside on the primary boot drive together with
the big-4 partitions). Even if the non-big-4 partitions
are NOT needed for the boot-up process, you are at least
ensuring that the next time you re-encapsulate this primary
root disk, you will have the SAME exact partitioning scheme.
7.B. "Non-rootdisk" partitions like data/application partitions.
(Example: /home, /apps, /oracle, /export/home all located
on a non-rootdisk, non-primary drive assuming that they
were ENCAPSULATED too!!!!). This however is NOT mandatory
and NOT required since for obvious reasons these partitions
are really NOT needed for the boot-up process under regular
Solaris hard partitions. What I am ensuring here is the fact
that you are at least ensuring reverting back to the original
hard partitions and know where the data previously was prior
to the encapsulation process. This ALSO works well in the
event that you split up your big-4 partitions on two separate
drives and encapsulated BOTH drives. Say, "/" and "/usr on
a drive called "c0t0d0" and "/var", "swap" and a data partition
called "/home" on a drive called "c0t1d0". Again, to be
identical for all drives, you recover ALL hard partitions
(big-4 obviously gets retained so no need to recover them)
even for non-primary, non-root disks.
Why do the "big-4" partitions get retained when you encapsulate
the boot disk? Because the encapsulation binary/script looks
for a hardcoded tag and flag such as:
PARTITION TAG MOUNT_POINT
0 2 /
2 5 backup <--- NEVER TOUCHED AS THIS IS
1 3 swap THE WHOLE DRIVE.
4 7 /var
6 4 /usr
And since this is labeled for the root disk, when you encapsulate
it, these Solaris hard partitions get retained and DO NOT get
CHANGED even after they become defined under Veritas Volume Manager
volumes.
7.A. Instructions For Recovering Hard PartitionsFfor "Non-big-4"
Partitions On The Primary Boot Disk:
-----------------------------------------------------------
Refer to /a/etc/vfstab.prevm (which you've converted to be your
/a/etc/vfstab) and /a/etc/vx/reconfig.d/disk.d/c#t#d#/vtoc
files to figure out what filesystems were mounted on the original
boot drive, and how large each partition was or what the hard
cylinder boundaries were for those partitions.
NOTES: - "vfstab.prevm" contains your original "vfstab" file before
VxVM took control of your boot disk.
- "vtoc" contains your original boot disk Solaris hard
partition table before VxVM was used to encapsulate it.
- c#t#d# is your boot disk device name say c0t0d0 as
a concrete example.
To recover the hard partitions for your Non-big-4 partitions
on your primary boot disk you have 3 options:
--------
OPTION 1: Recovering A Single Slice/Partition Via 'fmthard -d':
--------
# /usr/sbin/fmthard -d part:tag:flag:start:size /dev/rdsk/c#t#d#s2
where:
stripping off the "0x" and retaining the last number only
on the right side. Example 0x2 = 2 for the "tag" field.
stripping off the "0x2" and retaining the last two
numbers on the right side. Example 0x201 = 01 for the
"flag" field.
file corresponding to the starting sector number of
the partition.
corresponding to the number of sectors in the partition.
Example:
========
Here is an example of /a/etc/vx/reconfig.d/disk.d/c0t0d0/vtoc:
#THE PARTITIONING OF /dev/rdsk/c0t0d0s2 IS AS FOLLOWS :
#SLICE TAG FLAGS START SIZE
0 0x2 0x200 0 66080
1 0x3 0x201 66080 66080
2 0x5 0x201 0 1044960
3 0x0 0x200 132160 141120
4 0x0 0x200 273280 564480
5 0x0 0x200 837760 206080
6 0x0 0x200 0 0
7 0x0 0x200 0 0
In this example, let's say /opt (a Non-big-4 partition) used
to be on slice 4 (before being encapsulated) and needed to
be recovered (obtaining the hard cylinder boundaries), so we do:
# fmthard -d 4:0:00:273280:564480 /dev/rdsk/c0t0d0s2
This will partition slice 4, with the default flag and tag,
the starting block of 273280 in sectors and a total size
of 564480 blocks (also in sectors).
---------
OPTION 2: Recovering A Single Slice/Partition Via 'format' Utility:
---------
Use Solaris' "format" utility to repartition and recover the hard
partition for your Non-big-4 partition. The trick here is to
translate the
into cylinders so that they can be entered into the
format -> partition utility as starting and ending cylinders for that
particular partition.
The formula is:
where:
'prtvtoc' output:
Example:
========
# prtvtoc /dev/rdsk/c0t0d0s2
/dev/rdsk/c0t0d0s2 partition map
*
* Dimensions:
* 512 bytes/sector
* 80 sectors/track
* 19 tracks/cylinder
* 1520 sectors/cylinder
* 3500 cylinders
* 2733 accessible cylinders
In our example, the
and the
273280 / 1520 = 179 cylinders
564480 / 1520 = 371 cylinders
Using the "format" utility you can then repartition it by entering the
appropriate values:
# format
Choose the drive.
format> pa
partition> pr
partition> 4
Enter partition id tags: opt
Enter partition permission flags: wm
Enter new starting cylinder: 179
Enter partition size: 371c
---------
OPTION 3: Recovering The Entire Slice/Partition From The "vtoc"
--------- Of That Disk Via 'fmthard -s':
To set/change the entire "vtoc" of the disk, use:
# /usr/sbin/fmthard -s vtocfile
where:
part:tag:flag:start:size
Example:
========
In our example of /a/etc/vx/reconfig.d/disk.d/c0t0d0/vtoc:
#THE PARTITIONING OF /dev/rdsk/c0t0d0s2 IS AS FOLLOWS :
#SLICE TAG FLAGS START SIZE
0 0x2 0x200 0 66080
1 0x3 0x201 66080 66080
2 0x5 0x201 0 1044960
3 0x0 0x200 132160 141120
4 0x0 0x200 273280 564480
5 0x0 0x200 837760 206080
6 0x0 0x200 0 0
7 0x0 0x200 0 0
# cp /a/etc/vx/reconfig.d/disk.d/c0t0d0/vtoc /tmp/vtocfile
# vi /tmp/vtocfile
Edit the file and make it look like the following:
0 2 00 0 66080
1 3 01 66080 66080
2 5 01 0 1044960
3 0 00 132160 141120
4 0 00 273280 564480
5 0 00 837760 206080
6 0 00 0 0
7 0 00 0 0
7.B. Instructions For Recovering Hard Partitions For "Non-big-4"
Partitions On Non-rootdisk:
-----------------------------------------------------------
Same as Step 7.A. (You can use any of the three options listed)
assuming that you reference the correct "vtoc" file for the particular
drive. As long as the drive had been encapsulated, it should
have a corresponding "vtoc" file in:
/a/etc/vx/reconfig.d/disk.d/c#t#d#/vtoc
8. Using the 'format' utility, remove the public & private region
partitions of VxVM.
You must zero out these slices (usually the tag is a "-" and
the flag is a "wu").
First, verify whether the VxVM's private and public region partitions
are still there and intact via the 'prtvtoc' output:
Make sure pub/priv areas are there, because it shows the
tag values as numeric numbers 14 and 15.
14 is the public area.
15 is the private area .
As an example for a disk c3t2d0,
# prtvtoc /dev/rdsk/c3t2d0s2
* /dev/rdsk/c3t2d0s2 partition map
*
* Dimensions:
* 512 bytes/sector
* 80 sectors/track
* 7 tracks/cylinder
* 560 sectors/cylinder
* 2500 cylinders
* 1866 accessible cylinders
*
* Flags:
* 1: unmountable
* 10: read-only
*
* First Sector Last
* Partition Tag Flags Sector Count Sector Mount Directory
2 5 01 0 1044960
3 15 01 0 1120 1119
4 14 01 1120 1043840 1044959
Based on this output you can clearly see that tag 15 is VxVM's
private region which is located on slice 3 and tag 14 is VxVM's
public region which is located on slice 4.
Now using the 'format' utility fix and zero out these slices.
Again like I said, "format" usually lists the public and private
regions by having a "-" for the Tag and "wu" for the Flag.
Example:
========
# format
Choose the disk c3t2d0.
format> pa
partition> pr
Part Tag Flag Cylinders Size Blocks
3 - wu 0 - 1 0.55MB (2/0/0) 1120
4 - wu 2 - 1865 509.69MB (1864/0/0) 1043840
Make the Tag "unassigned" and the flag "wm".
partition> 3
Enter partition id tags: unassigned
Enter partition permission flags: wm
Enter new starting cylinder: 0
Enter partition size: 0
partition> 4
Enter partition id tags: unassigned
Enter partition permission flags: wm
Enter new starting cylinder: 0
Enter partition size: 0
It will NOW look like:
Part Tag Flag Cylinders Size Blocks
3 unassigned wm 0 (0/0/0) 0
4 unassigned wm 0 (0/0/0) 0
You have now effectively removed VxVM's public and private
region partitions.
9. Create a /a/etc/vx/reconfig.d/state.d/install-db file.
# touch /a/etc/vx/reconfig.d/state.d/install-db
WHY? Because IF the root disk contains mirrors, and the system
boots up, the mirrors will get resynced, corrupting the changes
we just made. The file is checked by VxVM rc scripts in
/etc/init.d and determines that if it exists it will disable
VxVM from starting and enabling itself.
10. Reboot the system.
# /usr/sbin/reboot
11. You will NOW be booted on regular Solaris hard partitions
with Veritas Volume Manager NOT running or disabled.
Verify via:
# df -k
You should see that all your devices are being mounted
as "/dev/dsk" as opposed to "/dev/vx/dsk"
Remove the /etc/vx/reconfig.d/state.d/install-db file
(and the "root-done" file if existent only!).
# rm /etc/vx/reconfig.d/state.d/install-db
# rm /etc/vx/reconfig.d/state.d/root-done
12. Determine which scenarios you have for your "rootdg" disk group
configuration.
SCENARIO A. You created the primary and the mirror disks to be the ONLY
disks residing in the "rootdg" disk group while all other
disks and volumes reside in different disk groups like
"datadg" for example.
SCENARIO B. You created the primary and the mirror disks, including all
other disks and volumes to ALL be in the "rootdg"
disk group
-----------
SCENARIO A:
-----------
If Scenario A., you will need to run 'vxinstall' and choose
to encapsulate your boot disk. In this scenario since the only
disks residing under rootdg are your boot and mirror drive it
is SAFE to run "vxinstall" and NOT worry about losing any data
or non-root volumes. This will re-create your rootdg disk group
and thus the need to be certain that only the boot and mirror
drives were the only disks part of this disk group.
The command is:
12.A.1 Run "vxinstall".
# /usr/sbin/vxinstall
- Choose 'Custom Install'.
- Choose 'Encapsulate Boot Drive c#t#d#'.
- Choose to 'Leave other disks alone', as it will go
thru all your other controllers as well assuming
that you did NOT create a '/etc/vx/disks.exlcude'
file that essentially disallows VxVM to touch the
disks defined within this file.
- Multiple reboots will then occur. Be sure you get
to the point where it says: "The system must be
shutdown... Shutdown now?" line.
You can "yes" to this or choose to initiate an
'init 6' or a 'reboot' command later on.
12.A.2 Once the reboot occurs what happens is:
- /etc/vfstab will be updated to reflect "/dev/vx" devices
defined under volumes for the boot partitions such as
(/, swap, /var, /usr, /opt)
- /etc/system will re-put the 2 entries previously removed:
rootdev:/pseudo/vxio@0:0
set vxio:vol_rootdev_is_volume=1
- /etc/vx/reconfig.d/state.d/install-db gets automatically
removed.
12.A.3 System will now be booted with VxVM started and enabled.
You can now verify this via a 'vxdctl mode' and the output
mode should say 'enabled'.
# vxdctl mode
Mode: enabled
(See DETAILED steps discussed in a later step, Step 14.).
-----------
SCENARIO B:
-----------
If Scenario B., you will need to go thru the so-called "E-38"
(Also known as Appendix B.8.24 in SEVM 2.6 Installation Guide)
Volume Recovery Procedure below:
E-38 Volume Recovery:
=====================
In recovering your Volume Manager configurations follow the
following steps:
12.B.1. Remove the /etc/vx/reconfig.d/state.d/install-db file
(and the "root-done" file if existent).
# rm -rf /etc/vx/reconfig.d/state.d/install-db
# rm -rf /etc/vx/reconfig.d/state.d/root-done
12.B.2. Start VxVM I/O daemons.
# vxiod set 10
12.B.3. Start the volume manager configuration daemon, vxconfigd
in disabled mode.
# vxconfigd -m disable
12.B.4. Initialize the vxconfigd daemon.
# vxdctl init
Note: In some cases, if you happen to have re-installed
your VxVM packages on a particular host or retained
the same packages on the same host BUT renamed the
hostname of the said host, you will have to run
'vxdctl init old_hostname' where:
old_hostname is obtained from the 'hostid' field
of a 'vxprivutil list' output. Since this entails
recovering your existing disks/volumes under the
diskgroup called "rootdg" AND other disks/volumes
on other disk groups, the command is run only
once. A reboot will then cause your /etc/vx/volboot
file to match your new hostname, so there should be
no cause for alarm in reverting to your old hostname.
Compare your 'hostid' field from your /etc/vx/volboot
to your /etc/hostname. They MUST match. If they do
not, follow example below:
Explicit Example:
=================
Obtain the slice number of VxVM's private region partition.
# prtvtoc /dev/rdsk/c3t2d0s2
* /dev/rdsk/c3t2d0s2 partition map
*
* Dimensions:
* 512 bytes/sector
* 80 sectors/track
* 7 tracks/cylinder
* 560 sectors/cylinder
* 2500 cylinders
* 1866 accessible cylinders
*
* Flags:
* 1: unmountable
* 10: read-only
*
* First Sector Last
* Partition Tag Flags Sector Count Sector Mount Directory
2 5 01 0 1044960
3 15 01 0 1120 1119
4 14 01 1120 1043840 1044959
Tag 15 (VxVM private region) is on slice 3.
# /usr/lib/vxvm/diag.d/vxprivutil list /dev/rdsk/c3t2d0s3
diskid: 941127537.11942.lucky-sw
group: name=richdg id=942324120.12495.lucky-sw
flags: private autoimport
hostid: lucky-sw
version: 2.1
iosize: 512
public: slice=4 offset=0 len=1043840
private: slice=3 offset=1 len=1119
The 'hostid' field for this disk is "lucky-sw".
Armed with this information, you will then run:
# vxdctl init lucky-sw
12.B.5. Enable the vxconfigd daemon.
# vxdctl enable
12.B.6. Verify that vxconfigd is enabled.
# vxdctl mode
mode: enabled
12.B.7 After performing the "E-38" procedure, verify that you see
all your Volume Manager volumes as well as disk groups via:
# vxprint -ht
# vxdisk list
OR you can open your 'vxva' GUI interface and look
at your volumes.
13. Perform "rootability" cleanup.
In order to perform "rootability" cleanup, the VxVM daemons
should be up and running. They however require that a default
"rootdg" disk group and at least one disk be present within that
disk group BEFORE they can be properly started/enabled!
-----------
SCENARIO A:
-----------
13.A. If SCENARIO A. is chosen, you are essentially rebuilding
your "rootdg" disk group so there is NO need to clean up
anything,
13.A.1 Run "vxinstall".
# /usr/sbin/vxinstall
- Choose 'Custom Install'.
- Choose 'Encapsulate Boot Drive c#t#d#'.
- Choose to 'Leave other disks alone', as it will go
thru all your other controllers as well assuming
that you did NOT create a '/etc/vx/disks.exlcude'
file that essentially disallows VxVM to touch the
disks defined within this file.
- Multiple reboots will then occur. Be sure you get
to the point where it says: "The system must be
shutdown... Shutdown now?" line.
You can "yes" to this or choose to initiate an
'init 6' or a 'reboot' command later on.
13.A.2 Once the reboot occurs what happens is:
- /etc/vfstab will be updated to reflect "/dev/vx" devices
defined under volumes for the boot partitions such as
(/, swap, /var, /usr, /opt)
- /etc/system will re-put the 2 entries previously removed:
rootdev:/pseudo/vxio@0:0
set vxio:vol_rootdev_is_volume=1
- /etc/vx/reconfig.d/state.d/install-db gets automatically
removed.
13.A.3 System will now be booted with VxVM started and enabled.
You can now verify this via a 'vxdctl mode' and the output
mode should say 'enabled'.
# vxdctl mode
Mode: enabled
(See DETAILED steps in the next step, Step 14.).
-----------
SCENARIO B:
-----------
13.B. If SCENARIO B. is chosen, you will have to perform a
"rootability" cleanup.
----
CLI:
----
13.B.1 Verify that "old" volumes for the root disk still
show up before the unencapsulation process was done.
# vxprint -ht
Look for volumes that used to be mounted on filesystems
that was reverted back to Solaris hard partitions. You
have to be aware that since you are performing
"unencapsulation" and reverting back to Solaris hard
partitions, any volumes that used to house these
filesystems become "hanging" or "unused" volumes that
appear on your disk group BUT have no functionality at all
or not associated to any filesystem at all.
Look for the following volumes (Listed below are the
most commonly used naming convention for the root drive
volumes):
- "rootvol" volume which contains the root filesystem.
- "swapvol" volume which contains the swap area.
- "usr" volume which contains the /usr filesystem.
- "var" volume which contains the /var filesystem.
- "opt" which contains the /opt filesystem.
13.B.2 Recursively remove these root drive volumes.
# vxedit -fr rm rootvol
# vxedit -fr rm swapvol
# vxedit -fr rm usr
# vxedit -fr rm var
# vxedit -fr rm opt
Note: Repeat the 'vxedit -fr' for any root or non-root
disk volumes that were unencapsulated or reverted
back to Solaris hard partitions. The "-fr" means
forcibly and recursively remove them since these
volumes become "hanging" volumes and will be
re-created afterwards via a "re-encapsulation"
process.
13.B.3 IS OPTIONAL!!!
13.B.3 You can choose to remove the root and mirror
Volume Manager disks if you intend to start from
scratch and have the ability to choose your own
root and mirror drives. In doing so, you will
have to RE-INITIALIZEe the 2 drives and put them
into the rootdg disk group once more.
# vxdisk list
Look for the Volume Manager disk name for the boot
and mirror disks.
In this example, let's say the primary boot drive
will be called "rootdisk" and the secondary mirror
drive will be called "disk01".
Remove them via:
# vxdg rmdisk rootdisk
# vxdg rmdisk disk01
----
GUI:
----
13.B.1 Bring up the "vxva" GUI.
# /opt/SUNWvxva/bin/vxva
13.B.2 Click and choose the "rootdg" disk group icon.
Choose all root disk volumes (the root disk's mirror is
reflected by the root disk volume having 2 plexes or a
label "V" indicating that it's a volume).
- Highlight rootvol, swapvol, usr var, opt volumes
Basic Ops -> Volume Operations -> Remove Volumes Recursively
Note: Repeat this for any root or non-root disk volumes
that were unencapsulated or reverted back to Solaris
hard partitions. This forcibly and recursively removes
them since these volumes become "hanging" volumes and
will be re-created afterwards via a "re-encapsulation"
process.
13.B.3 IS OPTIONAL!!!
13.B.3 You can choose to remove the root and mirror
Volume Manager disks if you intend to start from
scratch and have the ability to choose your own
root and mirror drives. In doing so, you will
have to RE-INITIALIZEe the 2 drives and put them
into the rootdg disk group once more.
If you choose to remove the root and mirror
Volume Manage disks, then do the following:
- Highlight the root and mirror Volume Manager disks
(Label is a "D").
Advanced Ops -> Disk Operations -> Remove
14. You have NOW SUCCESSFULLY UNENCAPSULATED your root disk!
Congratulations. You may now re-encapsulate and re-mirror
your root disk. Follow the steps below:
SCENARIO A.
In SCENARIO A., Re-encapulation of the boot disk is done via
'vxinstall' already discussed earlier. Here is the complete
walk-thru once more:
14.A.1 Run "vxinstall".
# /usr/sbin/vxinstall
- Choose 'Custom Install'.
- Choose 'Encapsulate Boot Drive c#t#d#'.
- Choose to 'Leave other disks alone', as it will go
thru all your other controllers as well assuming
that you did NOT create a '/etc/vx/disks.exlcude'
file that essentially disallows VxVM to touch the
disks defined within this file.
- Multiple reboots will then occur. Be sure you get
to the point where it says: "The system must be
shutdown... Shutdown now?" line.
You can "yes" to this or choose to initiate an
'init 6' or a 'reboot' command later on.
14.A.2 Once the reboot occurs what happens is:
- /etc/vfstab will be updated to reflect "/dev/vx" devices
defined under volumes for the boot partitions such as
(/, swap, /var, /usr, /opt)
- /etc/system will re-put the 2 entries previously removed:
rootdev:/pseudo/vxio@0:0
set vxio:vol_rootdev_is_volume=1
- /etc/vx/reconfig.d/state.d/install-db gets automatically
removed.
14.A.3 System will now be booted with VxVM started and enabled.
You can now verify this via a 'vxdctl mode' and the output
mode should say 'enabled'.
# vxdctl mode
Mode: enabled
14.A.4 Verify that you see your other disk groups (aside from rootdg)
and volumes under them via 'vxprint -ht', 'vxdisk list', or
'vxva' GUI.
# vxprint -ht
Look for the list of the disk groups, the volumes and
disks that were not part of the "rootdg" disk group.
They should show up including your existing "rootdg"
disk group and it's volumes and disks under it.
# vxdisk list
An example entry of a drive called "newdg" that was part
of a disk group called "newdg" being seen properly:
c3t1d0s2 sliced newdg03 newdg online
# vxva
Look at the GUI and you should see your icons for your
other disk groups. If you choose them, look at the
volumes and disks being seen within that particular
disk group.
14.A.5 Since a vxinstall was run, and we chose to encapsulate
your boot disk, typically the mirror disk will show up
as already being initialized but still not part of
any disk group and having no Volume Manager disk name
associated to it. For a better description this is what
it would look like for the mirror disk from a
'vxdisk list' output:
c#t#d#s2 sliced - - online
This indicates that 'c#t#d#s2' has been initialized but
is NOW NOT part of any disk group or has no Volume Manager
disk name.
Re-initialize or just add it back into the rootdg
disk group.
# vxdiskadm
- Choose option #1 (Add or initialize one or more disks).
- Choose the 'c#t#d#' for the mirror drive.
- It might then tell you that the drive has already
been initialized, and will ask you re-initialize?
You can either say "yes" or "no" to this. Just
answer "yes" to re-initialize it.
- It will then prompt you to enter your own Volume
Manager disk name for this drive or you can choose
the default name of "diskXX" (example disk01)
assigned by VxVM.
14.A.6 Re-mirror your root disk to this re-initialized drive
which will now become your mirror disk via:
# vxdiskadm
- Choose option #6 (Mirror volumes on a disk).
- Follow the prompts again as it is "user-friendly"
and will ask for the "source_drive" and "destination/mirror
drive". As an example, say a primary boot disk c0t0d0
called "rootdisk" being mirrored to a secondary mirror
disk called "disk01".
- 'vxdiskadm option #6' will do all the mirroring processes
for all the root disk volumes which are rootvol, swapvol,
usr, var and opt volumes. You DO NOT need to do each
volume manually and separately. This is the ease-of-use
and simplicity of using the 'vxdiskadm' utility.
14.A.7 Check to see if all other volumes from other disk groups
(non-rootdg) are "started" or not via:
# vxinfo
- A vxinfo with no flags will list out "root disk"
volumes only.
# vxinfo -g {DISK_GROUP_NAME.EN_US}
where {DISK_GROUP_NAME.EN_US} is "datadg" for example.
Mode should be: "started".
If not start up all your volumes via:
# vxvol startall
OR individually start them up via:
# vxvol start {VOLUME_NAME.EN_US}
where {VOLUME_NAME.EN_US} is "vol01" for example.
14.A.8 Lastly, edit your /etc/vfstab and put back all the mounts
for these volumes (/dev/vx) devices. You can manually
mount them too if you want to verify the integrity of the
data on these volumes which SHOULD ALL be GOOD.
- Edit vfstab using the "vi" editor:
As an example, adding a particular volume back into this
file:
/dev/vx/dsk/newdg/vol01 /dev/vx/rdsk/newdg/vol01 /home
ufs 3 yes -
Do this for all other volumes since the only entries you'll
have in here will be for your root disk volumes.
14.A.9 Re-mount all non-root-disk volumes (non-rootdg disk group).
# mountall
or manually mount them by hand one at a time via:
FOR UFS:
# mount /dev/vx/dsk/newdg/vol01 /home
FOR VXFS:
# mount -F vxfs /dev/vx/dsk/newdg/vol01 /home
or reboot the host via:
# reboot
Rebooting one last time ensures that everything is
in an "okay" state, "ENABLED ACTIVE" from a
'vxprint -ht' perspective and GUI normalcy from
the 'vxva' GUI.
Note: In the event that you are unable to mount the
volumes even after starting them, ensure that
you run 'fsck' on the volumes.
Example:
FOR UFS:
# fsck /dev/vx/rdsk/newdg/vol01
FOR VXFS:
# fsck -F vxfs -o full,nolog /dev/vx/rdsk/newdg/vol01
SCENARIO B.
In SCENARIO B., Re-encapulation is done via the 'vxdiskadm'
utility. (PLEASE REFERENCE INFODOC ID 15838 - HOW TO
ENCAPSULATE BOOT DISK USING 'VXDISKADM').
14.B.1 If you chose to REMOVE the Volume Manager disks
as well (following the optional step, Step 13.B.3
say, "rootdisk" for the primary drive and "disk01"
for the mirror which were removed via 'vxdg rmdisk rootdisk'
and 'vxdg rmdisk disk01'), then do the following:
First re-initialize and add the two drives back into the
rootdg disk group. As an EXAMPLE again, let's say the
primary drive c0t0d0 which will be called "rootdisk"
and the mirror drive c0t1d0 which will be called "disk01"
is to be intialized and added back into the rootdg disk
group. As this is an "EXAMPLE-ONLY" basis, choose the
appropriate c#t#d# and Volume Manager disk name or use
the defaults assigned by VxVM.
# vxdiskadm
- Choose option #1 (Add or initialize one or more disks)
and intialize the root and mirror disks.
- Follow the prompts and choose to initialize the c#t#d#
and give their VxVM disk names and tell vxdiskadm that
they are to be added into the rootdg disk group.
When in doubt which drives you want to initialize
always do a 'list' to list out the drives that can be
selected. In our example we should choose "c0t0d0"
and name it as "rootdisk" for the primary disk and
choose "c0t1d0" and nae this as "diks01" for the
mirror disk.
14.B.2 Verify that both disks now show up under the rootdg
disk group via 'vxdisk list' output.
# vxdisk list
EXAMPLE:
Primary c0t0d0 called "rootdisk.
Mirror c0t1d0 called "disk01".
OUTPUT WILL BE:
c0t0d0s2 sliced rootdisk rootdg online
c0t1d0s2 sliced disk01 rootdg online
Note: If the optional step, Step 13.B.3 was NOT done, then skip
steps 14.B1. and 14.B.2 and instead follow Step 14.B.3!!!
14.B.3 Re-encapsulated the root disk using the 'vxdiskadm'
utility.
# vxdiskadm
- Choose option #2 (Encapsulate one or more disks)
to re-encapsulate the root disk that was
previously initialized.
- Again follow the prompts as they should be
"user-friendly" and will direct you waht to choose
next.
14.B.4 Re-mirror the root disk to the designated mirror disk.
- Choose option #6 (Mirror volumes on a disk).
- Follow the prompts again as it is "user-friendly"
and will ask for the "source_drive" and "destination/mirror
drive". As an example, say a primary boot disk c0t0d0
called "rootdisk" being mirrored to a secondary mirror
disk called "disk01".
- 'vxdiskadm option #6' will do all the mirroring processes
for all the root disk volumes which are rootvol, swapvol,
usr, var and opt volumes. You DO NOT need to do each
volume manually and separately. This is the ease-of-use
and simplicity of using the 'vxdiskadm' utility.
14.B.5 Check to see if all other volumes from other disk groups
(non-rootdg) are "started" or not via:
# vxinfo
- A vxinfo with no flags will list out "root disk"
volumes only.
# vxinfo -g {DISK_GROUP_NAME.EN_US}
where {DISK_GROUP_NAME.EN_US} is "datadg" for example.
Mode should be: "started".
If the mode says "startable" or anything that
doesn't specifically say "started", then start
up all your volumes via:
# vxvol startall
OR individually start them up via:
# vxvol start {VOLUME_NAME.EN_US}
where {VOLUME_NAME.EN_US} is "vol01" for example.
14.B.6 Lastly, edit your /etc/vfstab and put back all the mounts
for these volumes (/dev/vx) devices. You can manually
mount them too if you want to verify the integrity of the
data on these volumes which SHOULD ALL be GOOD.
- Edit the /etc/vfstab using the "vi" editor:
As an example, adding a particular volume back into this
file:
/dev/vx/dsk/newdg/vol01 /dev/vx/rdsk/newdg/vol01 /home
ufs 3 yes -
Do this for all other volumes since the only entries you'll
have in here will be for your root disk volumes.
14.B.7 Re-mount all non-root-disk volumes (non-rootdg disk group).
# mountall
or manually mount them by hand one at a time via:
FOR UFS:
# mount /dev/vx/dsk/newdg/vol01 /home
FOR VXFS:
# mount -F vxfs /dev/vx/dsk/newdg/vol01 /home
or reboot the host via:
# reboot
Rebooting one last time ensures that everything is
in an "okay" state, "ENABLED ACTIVE" from a
'vxprint -ht' perspective and GUI normalcy from
the 'vxva' GUI.
Note: In the event that you are unable to mount the
volumes even after starting them, ensure that
you run 'fsck' on the volumes.
Example:
FOR UFS:
# fsck /dev/vx/rdsk/newdg/vol01
FOR VXFS:
# fsck -F vxfs -o full,nolog /dev/vx/rdsk/newdg/vol01
ENCAPSULATION PROCESS FOR A ROOT AND NON-ROOT DISK EXPLAINED:
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Encapsulating the Root Disk:
============================
Why encapsulate a root disk?
Two reasons:
1) to mirror it.
2) to have a disk in rootdg.
What do you need to encapsulate the root disk?
Two slices with no cylinders assigned to them.
Either:
a) Two cylinders that are not part
of any slice. These 2 cylinders must
be at the beginning or at the end of
the disk.
or
b) A 'swap' partition that VxVM can
take 2 cylinders away from.
Before and after 'format' outputs.
-----BEFORE----- --------AFTER--------
0 / 0-100 / 0-100
1 swap 101-200 swap 101-200
2 backup 0-502 backup 0-502
3 /usr 201-300 /usr 201-300
4 /opt 301-400 - 0
5 /var 401-500 /var 401-500
6
7
What's the difference between encapsulating root
vs non-root disks?
Non-root disk looses all hard partitions
Root disk retains 4 (the "big 4") hard
partitions, but all others are removed.
However, on both root & non-root disks, data is
never MOVED.
When you have a root-disk encapsulated:
- The /etc/vfstab file is modified to
mount volumes instead of hard partitions.
Below the mounts, it puts comments
containing the original partitions:
#Note: volume rootvol (/) encapsulated partition c0t3d0s0
#Note: volume swapvol (swap) encapsulated partition c0t3d0s1
#Note: volume opt (/opt) encapsulated partition c0t3d0s4
#Note: volume usr (/usr) encapsulated partition c0t3d0s5
#Note: volume var (/var) encapsulated partition c0t3d0s6
- Creates the /etc/vx/reconfig.d/state.d/root-done file
(not used for anything?)
- Creates a copy of the /etc/vfstab file in /etc/vfstab.prevm.
Warning: This file, once created, is never modified.
In other words, if new entries
are added to the 'real' /etc/vfstab file, this
one will be out of date.
- The /etc/system file contains two additional lines:
rootdev:/pseudo/vxio@0:0
set vxio:vol_rootdev_is_volume=1
- Even if you create an install-db file, the vxconfigd
will start up anyway! Why?
Because if root is under VxVM control, it HAS to. The
/etc/init startup scripts check the /etc/system file to
determine if root is encapsulated.
Encapsulating Non-root disks:
=============================
Why encapsulate? Many reasons...
- To mirror it.
- To grow it (concatenate).
- To "change" it to a striped volume.
Run 'vxdiskadm' (option #2) to encapsulate any disk
(either a root or non-root disk).
Why does it have to reboot afterwards?
- It has to repartition the drive.
- It has to change the /etc/vfstab file and must
mount the "vx" devices
(/dev/vx/dsk/rootdg/vol01) instead of the hard
partitions (/dev/dsk/c0t3d0s0).
What do you need to encapsulate?
- Two slices with no cylinders assigned to them.
- Two cylinders that are not part of any slice.
These 2 cylinders must be at the beginning or
at the end of the disk.
What happens AFTER you encapsulate a non-root disk?
Before:
# df -kl
Filesystem kbytes used avail capacity Mounted on
/dev/dsk/c1t1d4s0 96455 9 86806 0% /data0
/dev/dsk/c1t1d4s1 192423 9 173174 0% /data1
/dev/dsk/c1t1d4s3 288391 9 259552 0% /data3
/dev/dsk/c1t1d4s4 384847 9 346358 0% /data4
# format
Part Tag Flag Cylinders Size Blocks
0 unassigned wm 0 - 203 100.41MB (204/0/0)
1 unassigned wm 204 - 610 200.32MB (407/0/0)
2 backup wu 0 - 2035 1002.09MB (2036/0/0)
3 unassigned wm 611 - 1220 300.23MB (610/0/0)
4 unassigned wm 1221 - 2033 400.15MB (813/0/0)
5 unassigned wm 0 0 (0/0/0)
6 unassigned wm 0 0 (0/0/0)
7 unassigned wm 0 0 (0/0/0)
# prtvtoc /dev/rdsk/c1t1d4s2
* First Sector Last
* Partition Tag Flags Sector Count Sector Mount Directory
0 0 00 0 205632 205631 /data0
1 0 01 205632 410256 615887 /data1
2 0 00 0 2052288 2052287
3 0 01 615888 614880 1230767 /data3
4 0 01 1230768
Internals: The encapsulation process creates the following files:
# ll /etc/vx/reconfig.d/state.d
total 0
-rw-rw-rw- 1 root other 0 Feb 13 13:18 init-cap-part
-rw-rw-rw- 1 root other 0 Feb 13 13:18 reconfig
# more /etc/vx/reconfig.d/disks-cap-part
c1t1d4
# ls -l /etc/vx/reconfig.d/disk.d/c1t1d4
total 8
-rw-rw-rw- 1 root other 7 Feb 13 13:18 dmname
-rw-rw-rw- 1 root other 1143 Feb 13 13:18 newpart
-rw-rw-rw- 1 root other 476 Feb 13 13:18 vtoc
# more /etc/vx/reconfig.d/disk.d/c1t1d4/dmname
disk01
# more /etc/vx/reconfig.d/disk.d/c1t1d4/newpart
#volume manager partitioning for drive c1t1d4
0 0x0 0x000 0 0
1 0x0 0x000 0 0
2 0x0 0x200 0 2052288
3 0x0 0x000 0 0
4 0x0 0x000 0 0
5 0x0 0x000 0 0
6 0xe 0x201 0 2052288
7 0xf 0x201 2050272 2016
#vxmake vol data0 plex=data0-%%00 usetype=gen
#vxmake plex data0-%%00 sd=disk01-B0,disk01-%%00
#vxmake sd disk01-%%00 disk=disk01 offset=0 len=205631
#vxmake sd disk01-B0 disk=disk01 offset=2050271 len=1
putil0=Block0 comment="Remap of block 0
#vxvol start data0
#rename c1t1d4s0 data0
#vxmake vol data1 plex=data1-%%01 usetype=gen
#vxmake plex data1-%%01 sd=disk01-%%01
#vxmake sd disk01-%%01 disk=disk01 offset=205631 len=410256
#vxvol start data1
#rename c1t1d4s1 data1
#vxmake vol data3 plex=data3-%%02 usetype=gen
#vxmake plex data3-%%02 sd=disk01-%%02
#vxmake sd disk01-%%02 disk=disk01 offset=615887 len=614880
#vxvol start data3
#rename c1t1d4s3 data3
#vxmake vol data4 plex=data4-%%03 usetype=gen
#vxmake plex data4-%%03 sd=disk01-%%03
#vxmake sd disk01-%%03 disk=disk01 offset=1230767 len=819504
#vxvol start data4
#rename c1t1d4s4 data4
# more /etc/vx/reconfig.d/disk.d/c1t1d4/vtoc
#THE PARTITIONING OF /dev/rdsk/c1t1d4s2 IS AS FOLLOWS :
#SLICE TAG FLAGS START SIZE
0 0x0 0x200 0 205632
1 0x0 0x201 205632 410256
2 0x0 0x200 0 2052288
3 0x0 0x201 615888 614880
4 0x0 0x201 1230768
5 0x0 0x000 0 0
6 0x0 0x000 0 0
7 0x0 0x000 0 0
After Reboot:
After rebooting the system, you end up with four new volumes.
The names of the volumes will be "data0", "data1", "data2",
and "data3".
# df -kl
Filesystem kbytes used avail capacity Mounted on
/dev/vx/dsk/data0 96455 9 86806 0% /data0
/dev/vx/dsk/data1 192423 9 173174 0% /data1
/dev/vx/dsk/data3 288391 9 259552 0% /data3
/dev/vx/dsk/data4 384847 9 346358 0% /data4
# vxprint -ht
DM NAME DEVICE TYPE PRIVLEN PUBLEN STATE
V NAME USETYPE KSTATE STATE LENGTH READPOL PREFPLEX
PL NAME VOLUME KSTATE STATE LENGTH LAYOUT NCOL/WID MODE
SD NAME PLEX DISK DISKOFFS LENGTH [COL/]OFF DEVICE MODE
dm disk01 c1t1d4s2 sliced 2015 2050272 -
dm disk02 c1t0d4s2 sliced 2015 2050272 -
dm disk03 c1t3d4s2 sliced 2015 2050272 -
v data0 gen ENABLED ACTIVE 205632 ROUND -
pl data0-01 data0 ENABLED ACTIVE 205632 CONCAT - RW
sd disk01-B0 data0-01 disk01 2050271 1 0 c1t1d4 ENA
sd disk01-04 data0-01 disk01 0 205631 1 c1t1d4 ENA
v data1 gen ENABLED ACTIVE 410256 ROUND -
pl data1-01 data1 ENABLED ACTIVE 410256 CONCAT - RW
sd disk01-03 data1-01 disk01 205631 410256 0 c1t1d4 ENA
v data3 gen ENABLED ACTIVE 614880 ROUND -
pl data3-01 data3 ENABLED ACTIVE 614880 CONCAT - RW
sd disk01-02 data3-01 disk01 615887 614880 0 c1t1d4 ENA
v data4 gen ENABLED ACTIVE 819504 ROUND -
pl data4-01 data4 ENABLED ACTIVE 819504 CONCAT - RW
sd disk01-01 data4-01 disk01 1230767 819504 0 c1t1d4 ENA
# format
Part Tag Flag Cylinders Size Blocks
0 unassigned wm 0 0 (0/0/0)
1 unassigned wm 0 0 (0/0/0)
2 backup wm 0 - 2035 1002.09MB (2036/0/0)
3 unassigned wm 0 0 (0/0/0)
4 unassigned wm 0 0 (0/0/0)
5 unassigned wm 0 0 (0/0/0)
6 - wu 0 - 2035 1002.09MB (2036/0/0)
7 - wu 2034 - 2035 0.98MB (2/0/0)
# prtvtoc /dev/rdsk/c1t1d4s2
* First Sector Last
* Partition Tag Flags Sector Count Sector Mount Directory
2 0 00 0 2052288 2052287
6 14 01 0 2052288 2052287
7 15 01 2050272 2016 2052287
Issue/Introduction
