How to create a useable backup image before upgrade

[nimrod53]

Hi All,

I have a DS10 with a single disk and no external backup device attached. It is currently running openVMS 7.3-2 and I would like to upgrade it to V8.3.

I would like to do a standalone backup and also be able to restore it back to the working 7.3-2 in case the upgrade corrupts the disk.

I have 6 other standalone DS10s running V7.3-2 on the network.

The rest of the workstations also are single disk without any backup tape device.

I appreciate your comments and advice.

[Barry Alford]

If you can get hold of another SCSI disk, you can open up the case and connect it. Then boot from OVMS CD, select option 8 (DCL) and do an image backup:

$ mount/over=id
$ mount/foreign
$ backup/image

where:
is your original OVMS system disk
is the new disk

You will have to set a jumper on the to give it a different SCSI ID from the .

[Karl Rohwedder]

If you have enough free space on one of the other system's disk, you may backup your current system disk with
$ BACK/IMAG/IGN=INTER SYS$SYSDEVICE: node"user pass"::sys$sysdevice:[000000]backup.bck/SAVE
(perhaps another target directory...)

BUT: To restore your systemdisk in case of errors during upgrade you will have to connect the remote systemdisk to your local system.
AND: be careful not to fill up the remote disk, which may lead to problems

regards Kalle

[nimrod53]

Unfortunately, I don't have a spare disk.

[Karl Rohwedder]

An alternative would be a cheap IDE disk temporarily connected to the DS10 (you have to open the cover).

regards Kalle

[nimrod53]

I won't be able to copy the image backup over to the corrupted one. So I was thinking of doing a network cluster with the corrupted DS10 as the slave booting from the one with the image backup, then do a restore to the corrupted disk. Will this strategy work?

[Joseph Huber_1 · Accepted solution]

>> cluster with the corrupted DS10 as the slave booting from the one with the image backup, then do a restore to the corrupted disk.

Yes, if You have a cluster license, then of course this is the simplest thing.

In this case, and if the space for an image backup to a host disk is limited, then You could only backup the non-system files, hoping You have them separated in their own directories. (Assuming You have the same system configuration on all cluster members).

Do the clustering before doing the backup, then no DECnet is needed to copy the saveset to the host-system, just mount the system disk of the cluster-member, then use it in the saveset spec.

[comarow]

Please let me hit it from the perspective of managing your management.

Part of systems management is getting your management to do what is essential.
Document that in these days of disastor recover, a "backout recovery option" protect the company from a disastor. But, selling it always remains somehow getting management to think it was their idea.

Face it, modern disks are some cheap. If it were my site, much more expensive than down time.

Regardless of this situation, it exposes a real problem, how to have the ability to restore a disk, not only from this node, but all your nodes. Thus, a solution could
work for all nodes. Lay out the math. However, somehow leave the obvious solution out there for them to draw that solution, and then be prepared to say how great their idea is.

You can create save sets on a remote node using Decnet. In fact, you can restore a disk from an image backup stored on a remote node to a local system. What you can't do over the network is do an image backup from disk to disk.

IE: You can do a backup/image mydisk: othernode"username password"othernode::disk:[directory]savesetname.sav/save

You can also revere that.

Backup/image othernode::disk:[dir]savesetmae/sav

Now restoring it will require another step. You have the saveset on your disk.
Just a bootable disk for repair purposes.

Obviously, an external disk and your sneakers will provide a solution for all your systems.

If you could cluster them, or even have an extra cluster license, the problems become much easire. Of course, you would need to set it up so the node can do boot into the cluster without a disk.

Lets face it, there days pulling a disk and moving it to a systems does 90% of this for you.

Have fun.... Bob - no longer HP, but a Deccie always in Spirit.

[Colin Butcher]

Assuming the machines to have SCSI discs and an external SCSI connector then just buy a 2nd user tape drive and use that to write an image backup to, booting from the VMS distribution CD.

Alternatively you could by an external storage shelf and write the backups to a disc.

Running that many machines with no backup device at all for any of them is not a good thing.

Cheers, Colin.

[nimrod53]

Hi All,

Many thanks for your responses and thoughts/concerns.

These DS10s are currently off production mode and will be used as cold backup systems. They were previously in prodcution but has been replaced by Pcs running WindowsXP..boring..8-)

I guess the cheapest way to skin this cat in the event of a disaster is to reinstall 7.3-2 from CD and copy the image backup saveset file back to the root dir of the disk. Then bring the system with the installation CD and choose option 7 to restore the image backup files. Would this be a viable solution?

[Karl Rohwedder]

No,

when restoring an image backup, the target device is mounted /FOREIGN and connot be identical to the source device.

regards Kalle

[nimrod53]

thks for the heads up. Haven't been doing this for ages...

[nimrod53]

To make it simple, I'll juz copy the following files elsewhere - sysuaf.dat,rightslist.dat and netproxy.dat and the user directories.