Hp-UX backup to another station

[Leonardo Agopian]

Hello,

I administer a HP9000 A4323A, B180L server, HP-UX 10.20 and periodically we have full backup (fbackup) of the entire server to a DAT Tape. This server has external SCSI Disks (HP Disk System 2100). These are four 18.2GB hard disks, with mirroring enabled. There are also two volume groups created, with some logical volumes in each. This Disks are external, Single-Ended SCSI-2 (50 pins) and configuration is:

8/16/5.0.0 disk vg01 /dev/dsk/c1t0d0
8/16/5.2.0 disk vg01 /dev/dsk/c1t2d0
8/16/5.4.0 disk vg01 /dev/dsk/c1t4d0
8/16/5.6.0 disk vg01 /dev/dsk/c1t6d0
8/16/5.7.0 ct1 Initiator

Now I need to restore some information from November 2003, and this has to be made on another server because the original server is in production and I can't shut it down.

The problem is: Although I have a second server (exactly the same), I DON'T have a second mirroring system for restore to be done. As I need to do a full restore (frecover) I installed a HP-UX 10.20 on two hard disk, and I created THE SAME LVM configuration: two Volume Groups with corresponding Logical Volumes in each one.
The only difference is Hard Disks I'm installing are on another SCSI Bus:

Wide-Single Ended Ultra SCSI-3

with SCSI IDs 6 and 0 (one for each VG that exists on DAT backup)

After installing whole HP-UX and doing a full restore, I restarted station and these error messages appear:

couldn't query phisical volume '/dev/dsk/c1t6d0'
couldn't query phisical volume '/dev/dsk/c1t4d0'
couldn't query phisical volume '/dev/dsk/c1t0d0'
couldn't query phisical volume '/dev/dsk/c1t2d0'

After this, try to boot from SCSI ID #6, and it can't. Even a console can start up, so system crash and I can't do anymore.

First time I get this error, I didn't check SCSI Device Files on server to be restored.

I installed again HP-UX 10.20 and before doing restore, I check /dev/dsk with following files:

/dev/dsk/c0t6d0
/dev/dsk/c0t0d0

I think the problem is: "I'M USING ANOTHER BUS THAT WERE BACKED UP".
But is it the ONLY problem? Or perhaps there's no way to do a recover from a mirrored system to another NOT-MIRRORED?
What should I do? Should I change bus type? Or do I need other two hard disks, and let the "same" hardware configuration with four hard disks?
Another thing I don't know what is:

ct1 Initiator

as the ID#7 in HP /dev/dsk at Production Server configuration.

Another one: can I change a "card" for a disk? Perhaps if I can change from c0t6d0 to c1t6d0, this could work without hardware configuration.

The last: how can I check which hard disk is mirrored with wich? I mean, I want to know exactly what ID is mirrored with what ID.


Any suggestions accepted.
If more info is needed to analyze my problem, please let me know, I'll respond ASAP.

Best Regards,

Leonardo G. Agopian

[john kingsley]

I think the problem is that when you did the full restore, your vg00 LVM information (/etc/lvmtab) on server-2 was overwritten by the LVM information from server-1. You could try saving your original lvmtab file and then moving back in place after the restore. I'm not sure that this will work though.

Is it possible for you to restore only the information you need without restore the device information? You could try setting up an exclude file for the frecover, and have frecover skip over the device files.

Another approach could be to backup vg00 using vgcfgbcackup and then restore the original configuration using vgcfgrestore. You would probably have to run these in lvm maintenance mode.

You can find out which hard disk hold the mirrored copies of your lvols by running:
lvdisplay -v /dev/vg00/lvol#
for each of your lvols. This will show the physical volume for the original and mirrored copy for each extent.

Good Luck.

[Leonardo Agopian]

1) Do you think the only file I need to save and restore after frecover has been done is lvmtab? Or I need something else.

2) If I exclude device files, what files should I exclude? All in /dev ? Will work server after frecover?

3) What exactly backup vgcfgbackup? If it's the same in lvmtab, I think is the same try as in 1)

4) About finding out mirrored copies, lvdisplay -v was ok.

Thank you

[Darren Prior · Accepted solution]

Hi Leonardo,

As John mentioned above, you shouldn't recover the whole system using frecover. In the future it would be better to consider using Ignite to backup the whole system - it's then straightforward to clone the system onto your second server. From there you could work with the files in your test environment.

In your current situation I'd suggest you recover only the data and application files that you require - and leave all the operating system files alone.

best regards,

Darren.

[john kingsley]

1. I would start with lvmtab. This file contains a list of device files for each volume group. I think this is where your error is coming from. However, there is a good chance that you will hit problems with other files when you get past this point.

2. I would start by excluding everything under /dev and /etc/lmtab. You will probably discover that there are other files under /etc that may need to be skipped as well. You may want to consider excluding /etc as well, and then restoring it separatly to another locatation, and then copying over the files that you need.

3. I would try vgcfgbackup/vgcfgrestore as a last resort. This command save your LVM VG configuration to a file. By default this command is run everything the LVM configuration is changed, and the backup files are saved in /etc/lmconf/vg##.conf. Using this approach to recover the vg00 configuration will be pretty tricky. You would need to boot into lvm maintanence mode (from ISL prompt type "hpux -lm")

4. You could consider another approach. As Darren mentions, the best way to do a full system backup and recovery is to create a make_recovery_tape. This command comes with a HP Ignite-UX, and is available for free on their web site. If you are not interested in the system data from backup tape, but are only really needing to clone your current server, and then restore user/application data from Nov. 2003. You could run make_tape_recovery on your production server, and then use the recovery_tape to build your second server. Then after it's built, you could do a frecover on only the pieces you need from the November backup. Since the servers have disks on different buses, you will need to preview the make_tape_recovery

Run: make_recovery -A -p

Then edit the preview file (/var/opt/ignite/recovery/config.recover) change the hardware paths for your disks.

Then run: make_recovery -r

This will take a while. Depending on the size of the system, it could take severy hours to build.

[Leonardo Agopian]

Although I couldn't check other options you gave me, I consider points for both of you, who was interested in solve my problem.

This is what I did:

1) After installing fresh HP-UX, I made a tar of the following files:

/etc/lvmtab
/etc/ioconfig
/etc/lvmconf (entire)
/stand/ioconfig
/dev/vg00 (entire)
/dev/vg01 (entire)

2) I did a frecover.
3) I tar again to original location.
Reboot the system

After that, everything was ok, with all applications I needed to obtain database information.

Other solutions about restore only part of applications I needed..... well.... there are many applications, and many interaction with Operating System, and I'm not sure what could I exclude for this to put it to run.
Indeed, I'll check Ignite software, for a faster backup-restore of Operating System.
Again, Thank you for your help.