Using alt_disk_copy

Cloning rootvg

David Tansley

Most system administrators have experienced the following scenario:

• A failed ML upgrade.
• It's getting to the end of the day.
• You cannot fix it.
• It's too late to get it resolved by third-party support.
• You need to back out.

Typically, this situation requires a rootvg restore, whether it uses a tape mksysb restore or a network boot restore. There is no doubt it is a pain! Using the alt_disk_copy method to take a copy of your rootvg only requires the time it takes to do a reboot to recover your rootvg to the pre-upgrade event. This article demonstrates how to implement alt_disk_copy when applying an AIX upgrade and how to recover rootvg. alt_disk_copy can also be used for testing two different versions of AIX. You simply upgrade one disk then boot off it, and when you need to go back to the other version, simply boot off that disk instead. Indeed, the alt_disk_copy is often used to clone the rootvg to a spare disk for regular on-line backup of rootvg. It can also be used as a hardware migration tool of rootvg.

This article focuses on a typical rootvg two-disk software mirror set-up. However, alt_disk_copy is not restricted to this two-disk set-up; the same principles apply to multiple software mirroring situations.

The alt_disk utilities consist of the following commands:

• alt_disk_copy performs disk cloning.
• alt_rootvg_op performs maintenance operations on the clone rootvg.
• alt_disk_mysysb performs a mksysb copy.

This demonstration does not discuss alt_disk_mysysb.
The filesets required for the alt commands are:

1 2 3

bos.alt_disk_install.boot_images bos.alt_disk_install.rte     bos.msg.en_US.alt_disk_install.rte

Overview information

Because the alt_disk_copy command takes a copy of the current running rootvg to another disk, be sure to have all the file systems mounted that you want cloned across. alt_disk_copy only copies the currently mounted file systems in rootvg. There is no need to stop processes to execute alt_disk_copy; however, this process can take some time, so it is best to do it at lunchtime or in the evening (remember it is taking a running copy). Once the copy has completed, you will be presented with two rootvg volume groups:

1 2	rootvg altinst_rootvg

where altinst_rootvg is the cloned non-active/varied off rootvg. The cloned rootvg has all its logical volumes prefixed with the name 'alt'. The boot list is also changed to boot off altinst_rootvg. AIX likes to do things like this; it assumes you will want to boot off the cloned and not the real rootvg. If the system is now rebooted and when the system comes back up, the original rootvg will become:

1	old_rootvg

The original altinst_rootvg becomes:

1

rootvg

If you decide to reboot off the old_rootvg, when the system comes back up, the old_rootvg becomes:

1

rootvg

The rootvg becomes:

1	altinst_rootvg

Do not worry about the renaming of the original and cloned rootvg. I will demonstrate this shortly.
With a successful completion of an upgrade, the disk containing the cloned rootvg can then be destroyed using the alt_rootvg_op and mirrored back in. If the upgrade event has gone disastrously, there is no real problem--simply take a snapshot for third-party support, then boot off the good rootvg. For users to log in, it is business as normal.
When you get a response back from support on the fix, during off-line hours, simply reboot off the cloned rootvg and fix the issue. There is no need to go through the time-consuming tasks of re-applying the upgrade because you already have it on the cloned rootvg. Get the upgrade tested, and if it is all OK, destroy the cloned rootvg and mirror back in.

Do not use importvg or exportvg on the clone rootvg; use the alt commands instead.

With the cloned rootvg, you can mount the file systems by waking up the disk using alt_rootvg_op. Doing whatever works is required on the cloned file systems, and one would assume here to fix a patch of link, or gather information for third-party support, then put the disk back to sleep, which will also unmount the file systems.

Excluding directories when cloning

When cloning, you can exclude certain directories by creating the file: /etc/exclude.rootvg. The entries should start with the ^ /. characters. The '^' means to search for the string at the beginning of the line and the './' means relative to the current directory. You are advised to do this so alt_disk_copy does not misinterpret the command, as it uses grep to search for the string. So, make sure you provide the full pathname, prefixed with '^.' , for example, to exclude the following directories:

1 2	/home/reps /opt/installs

I could insert into the /etc/exclude.rootvg file:

1 2	^./home/reps ^./opt/installs

Make sure there are no empty lines after the last entry.

Let's get cloned!

Let's now go through a typical clone. Assume you have a software two-disk (hdisk0 and hdisk1) mirror of rootvg, and further assume that you are going to do a ML (or application upgrade, assuming it is installed in rootvg) upgrade on this system. I will demonstrate one way this can be done to clone the disk and after a successful upgrade will bring the disk back into rootvg and re-mirror. I will also demonstrate the actions you can take if the upgrade fails.

Pre-checks

Before unmirroring the rootvg, first take some time to ensure you are correctly mirrored and have no stale LV's, because if you do, the unmirrorvg will fail. Of course, you could always do a migratepv to move the missing LV's across if the unmirrorvg fails. A simple method to check that you are mirroring is to issue the command:

1	lsvg -l rootvg

For each row of data output, check that the output of the PPs column is double that of the LPs column.
Another method to check to see if you are mirroring is to use: lspv -l <hdiskx> and compare the output to make sure you have entries for each LV on both disks.
Next, issue the bosboot command. I personally always do this prior to either rebooting or disk operations involving rootvg; it is a good habit to have:

1 2	# bosboot -a bosboot: Boot image is 35803 512 byte blocks.

A listing of the disks being used for this demonstration is as follows:

1 2 3

# lspv hdisk0          0041a97b0622ef7f                    rootvg          active hdisk1          00452f0b2b1ec84c                    rootvg          active

Next, unmirror rootvg and take the disk that is going to be used for cloning out of rootvg. This demonstration uses hdisk1 to clone rootvg, so issue the unmirrorvg command:

1 2 3 4 5 6 7 8 9 10

# unmirrorvg rootvg hdisk1 0516-1246 rmlvcopy: If hd5 is the boot logical volume, please run 'chpv -c <disk name>'         as root user to clear the boot record and avoid a potential boot         off an old boot image that may reside on the disk from which this         logical volume is moved/removed. 0516-1804 chvg: The quorum change takes effect immediately. 0516-1144 unmirrorvg: rootvg successfully unmirrored, user should perform         bosboot of system to reinitialize boot records.  Then, user must modify         bootlist to just include:  hdisk0.

Next, take hdisk1 out of rootvg in readiness for the cloning:

1	# reducevg rootvg hdisk1

Confirm that the disk is now not assigned to any volume groups:

1 2 3

# lspv hdisk0          0041a97b0622ef7f                    rootvg          active hdisk1          00452f0b2b1ec84c                    None

Running alt_disk_copy

Now you are ready to issue the alt_disk_copy. Simply supply hdisk1 as a parameter to the command. The basic format is:

1	alt_disk_copy -d <hdisk to clone rootvg>

To use an exclude list, the basic format is:

1	alt_disk_copy -e /etc/exclude.rootvg -d <hdisk to clone rootvg>

The following output from the alt_disk_copy command has been truncated:

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

# alt_disk_copy -d hdisk1 Calling mkszfile to create new /image.data file. Checking disk sizes. Creating cloned rootvg volume group and associated logical volumes. Creating logical volume alt_hd5 Creating logical volume alt_hd6 Creating logical volume alt_hd8 Creating logical volume alt_hd4 Creating logical volume alt_hd2 Creating logical volume alt_hd9var Creating logical volume alt_hd3 Creating logical volume alt_hd1 Creating logical volume alt_hd10opt Creating /alt_inst/ file system. Creating /alt_inst/home file system. Creating /alt_inst/opt file system. Creating /alt_inst/tmp file system. …...... …...... for backup and restore into the alternate file system... Backing-up the rootvg files and restoring them to the alternate file system... Modifying ODM on cloned disk. Building boot image on cloned disk. forced unmount of /alt_inst/var forced unmount of /alt_inst/usr forced unmount of /alt_inst/tmp forced unmount of /alt_inst/opt forced unmount of /alt_inst/home ….. ….. Changing logical volume names in volume group descriptor area. Fixing LV control blocks... Fixing file system superblocks... Bootlist is set to the boot disk: hdisk1

At this stage, you now have a cloned rootvg called altinst_rootvg. Notice in the previous output alt_disk_copy has changed the bootlist to boot off the cloned rootvg, which is now hdisk1.

1 2 3

# lspv hdisk0          0041a97b0622ef7f                    rootvg          active hdisk1          00452f0b2b1ec84c                    altinst_rootvg

This can be confirmed by issuing the bootlist command:

1 2	# bootlist -m normal -o hdisk1 blv=hd5

At this point the ML upgrade can now be installed. After an ML upgrade you will need to reboot the system. For this demonstration, the ML upgrade will be installed on the real rootvg (that is hdisk0), so you need to change the bootlist now, because you want the system to come up with the new upgrade running.

1	# bootlist -m normal hdisk0

Confirm the change of the bootlist:

1 2	# bootlist -m normal -o hdisk0 blv=hd5

Next, install the ML upgrade, then reboot. After rebooting, the system presents the following rootvg and cloned rootvg. As can be seen, no root volume group has been renamed, because we booted off the real rootvg (hdisk0):

1 2 3

# lspv hdisk0          0041a97b0622ef7f                    rootvg          active hdisk1          00452f0b2b1ec84c                    altinst_rootvg

Next let's assume everything has gone OK on the upgrade and support users and the systems administrator has signed it off with no issues found. The alt_disk_copy can now be destroyed, and the disk brought back into rootvg for mirroring. Use the alt_rootvg_op command with the X parameter to destroy the cloned rootvg. The basic format is:

1 2 3 4

alt_rootvg_op -X < cloned rootvg to destroy> # alt_rootvg_op -X altinst_rootvg Bootlist is set to the boot disk: hdisk0

Next, extend rootvg to bring hdisk1, and then mirror up the disk:

1 2 3 4 5 6

# extendvg -f rootvg hdisk1 # mirrorvg rootvg hdisk1 0516-1804 chvg: The quorum change takes effect immediately. 0516-1126 mirrorvg: rootvg successfully mirrored, user should perform         bosboot of system to initialize boot records.  Then, user must modify         bootlist to include:  hdisk0 hdisk1.

Change the bootlist to include both disks and run bosboot:

1 2 3 4 5 6 7 8

# bootlist -m normal -o hdisk0 hdisk1 hdisk0 blv=hd5 hdisk1 # bosboot -a bosboot: Boot image is 35803 512 byte blocks. # bootlist -m normal -o hdisk0 blv=hd5 hdisk1 blv=hd5

For this demonstration, that's it: mission accomplished. The pgrade is installed with no issues. The system is operational. That's pretty much how alt_disk_copy works if all goes OK. But what if the upgrade fails? What options do you have? Let's look at that next.

Recovery positions, please

Let's now assume you have just installed the ML upgrade and rebooted, and issues have been found with the operational running of AIX. Remember, you currently have the disks in the following state:

1 2 3

# lspv hdisk0          0041a97b0622ef7f                    rootvg          active hdisk1          00452f0b2b1ec84c                    altinst_rootvg

At this point, a snapshot should be taken of the running system, in readiness for third-party support, for the call that you will undoubtedly log. Taking stock of the current situation, you have:

• rootvg: with post-upgrade issues.
• altinst_rootvg : with good copy pre-upgrade.

Take me back

To get back to the pre-upgrade, simply change the bootlist to boot off the (altinst_rootvg) hdisk1, then reboot. It's that simple:

1 2 3 4 5

# bootlist -m normal -o hdisk1 hdisk1 blv=hd5 # bootlist -m normal -o hdisk1 blv=hd5 # shutdown -Fr

After the reboot, you will be presented with the following rootvg disks:

1 2 3

# lspv hdisk0          0041a97b0622ef7f                    old_rootvg hdisk1          00452f0b2b1ec84c                    rootvg          active

Next, issue a bosboot and confirm the bootlist:

1 2 3 4

# bosboot -a bosboot: Boot image is 35803 512 byte blocks. # bootlist -m normal -o hdisk1 blv=hd5

The system is now back to the pre-upgrade state.

Post upgrade fixing

At a convenient time schedule that is agreed-upon with the end users, and with information provided by third-party support, you can then boot off the ML failed upgraded disk (hdisk0) and apply a fix that might solve the issue, so change the bootlist to boot off (old_rootvg) hdisk0 and reboot:

1 2	# bootlist -m normal -o hdisk0 # shutdown -Fr

After the reboot, in readiness to apply the fix, you will be presented with the following rootvg disks:

1 2 3

# lspv hdisk0          0041a97b0622ef7f                    rootvg          active hdisk1          00452f0b2b1ec84c                    altinst_rootvg

Next, apply the fix or instructions on how to fix it have been carried out, and assume the system is now operational again.
After the system has been tested and signed off bring in hdisk1, use the commands described earlier:
#

1	alt_rootvg_op -X altinst_rootvg

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Bootlist is set to the boot disk: hdisk0 # extendvg -f rootvg hdisk1 # mirrorvg rootvg hdisk1 bootlist -m normal -o hdisk0 hdisk1 hdisk0 blv=hd5 hdisk1 # bosboot -a bosboot: Boot image is 35803 512 byte blocks. #  bootlist -m normal -o hdisk0 blv=hd5 hdisk1 blv=hd5 # lspv hdisk0          0041a97b0622ef7f                    rootvg          active hdisk1          00452f0b2b1ec84c                    rootvg          active

Waking the disk up

Within a cloned rootvg environment, you can wake up the cloned rootvg to be active. All cloned file systems from the cloned rootvg will be mounted. It is quite useful because you have a good running system, but at the same time mount the file systems from the cloned rootvg for further investigation or file modification. When a cloned rootvg is woken up, it is renamed to:

1	altinst_rootvg

Do not issue a reboot while the cloned rootvg filesystems are still mounted, because unexpected results can occur. You can also rename a cloned rootvg, which is useful when you have more than one cloned rootvg.
Assume you have the disks in the following state:

1 2 3

# lspv hdisk0          0041a97b0622ef7f                    old_rootvg hdisk1          00452f0b2b1ec84c                    rootvg          active

To wake up a disk, the basic format is:

1	alt_rootvg_op -W -d < hdisk>

Let's now wake up old_rootvg (hdisk0):

1 2	# alt_rootvg_op -W -d hdisk0 Waking up old_rootvg volume group ...

Checking the state of the disks, you can see the old_rootvg has been renamed to altinst_rootvg and is now active.

1 2 3

# lspv hdisk0          0041a97b0622ef7f                    altinst_rootvg  active hdisk1          00452f0b2b1ec84c                    rootvg          active

The cloned file systems have been mounted, with the prefix of /alt_:

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

# df -m Filesystem    MB blocks      Free %Used    Iused %Iused Mounted on /dev/hd4         128.00    102.31   21%     2659    11% / /dev/hd2        1968.00    111.64   95%    40407    58% /usr /dev/hd9var      112.00     77.82   31%      485     3% /var /dev/hd3          96.00     69.88   28%      330     3% /tmp /dev/hd1         208.00    118.27   44%     1987     7% /home /proc                 -         -    -         -     -  /proc /dev/hd10opt    1712.00   1445.83   16%     6984     3% /opt /dev/alt_hd4     128.00    102.16   21%     2645    11% /alt_inst /dev/alt_hd1     208.00     33.64   84%     1987    21% /alt_inst/home /dev/alt_hd10opt   1712.00   1445.77   16%     6984     3% /alt_inst/opt /dev/alt_hd3      96.00     72.38   25%      335     2% /alt_inst/tmp /dev/alt_hd2    1968.00    100.32   95%    40407    59% /alt_inst/usr /dev/alt_hd9var    112.00     77.53   31%      477     3% /alt_inst/var

At this point file modification or further investigation can be carried out on the cloned rootvg. Now you can access the cloned file systems. Once these tasks have been carried out, put the cloned rootvg to sleep and in the same operation issue a bosboot on that disk. The basic format of the command is:

1	alt_rootvg_op -S -t <hdisk>

Let's now put the altinst_rootvg to sleep:

1 2 3 4 5 6 7 8 9 10 11 12

# alt_rootvg_op -S -t hdisk0 Putting volume group altinst_rootvg to sleep ... Building boot image on cloned disk. forced unmount of /alt_inst/var forced unmount of /alt_inst/usr forced unmount of /alt_inst/tmp forced unmount of /alt_inst/opt forced unmount of /alt_inst/home forced unmount of /alt_inst forced unmount of /alt_inst Fixing LV control blocks... Fixing file system superblocks...

The current state of the disks is now:

1 2 3

# lspv hdisk0          0041a97b0622ef7f                    altinst_rootvg hdisk1          00452f0b2b1ec84c                    rootvg          active

From the above demonstration, you can see the cloned rootvg name stayed the same: altinst_rootvg.
It is sometimes good to go back to the original state of the disks to save confusion, especially if you have more than one cloned disk. So rename altinst_rootvg back to old_rootvg. The basic format is:

1	alt_rootvg_op -v <new cloned rootvg name> -d <hdisk>

So in this example, you would issue:

1 2 3 4

# alt_rootvg_op -v old_rootvg -d hdisk0 # lspv hdisk0          0041a97b0622ef7f                    old_rootvg hdisk1          00452f0b2b1ec84c                    rootvg          active

Of course, you could rename the cloned rootvg to something more meaningful, if so desired.

1 2 3 4

# alt_rootvg_op -v bad_rootvg -d hdisk0 bash-2.05a# lspv hdisk0          0041a97b0622ef7f                    bad_rootvg hdisk1          00452f0b2b1ec84c                    rootvg          active

You cannot rename a cloned rootvg to altinst_rootvg; it is a reserved name.

From this point, the system is now operational or not, depending on the success of the fix, using the commands described earlier.
If the fix worked on (old_rootvg) hdisk0, then run with the new ML version.
Confirm that the disk will boot off hdisk0:

1	# bootlist -m normal -o hdisk0

Reboot:

1	# shutdown -Fr

Destroy the newly cloned disk (we rebooted off old_rootvg; it now becomes altinst_rootvg) hdisk1:

1	# alt_rootvg_op -X altinst_rootvg

Bring in hdisk1 into rootvg for mirroring:

1 2 3 4

# extendvg -f rootvg hdisk1 # mirrorvg rootvg hdisk1 # bosboot -a # bootlist -m normal -o hdisk0 hdisk1

If the fix did not work, then stay at the same ML version, and fix another day:
Confirm that the disk will boot off hdisk1:

1	# bootlist -m normal -o hdisk1

Destroy cloned disk (old_rootvg) hdisk0:

1	# alt_rootvg_op -X old_rootvg

Bring in hdisk0 into rootvg for mirroring:

1 2 3 4

# extendvg -f rootvg hdisk0 # mirrorvg rootvg hdisk0 # bosboot -a # bootlist -m normal -o hdisk0 hdisk1

Conclusion

This article showed that using the alt command is a quick way to recover rootvg if events go wrong on an AIX upgrade and how to mount the cloned rootvg filesystem on a running system. The alt command can also provide a path to a migration of a rootvg disk to another hardware. It is also very useful in having two different versions of AIX installed for testing migrating procedures.