Mirroring smaller disk to larger disk

[Ross McWalter]

Hi there -

We have 2 identical rp3440s (A and B) one with mirrored 73 Gb drives and another with x2 146 Gb. I was wondering whether there are any limitations in the following -

1. Take hot-swap 73 Gb drive from rp3440 A
2. Remove 146 Gb drive from rp3340 B
3. Plug in 73 Gb drive from rp3440 A into rp3440 B
4. Mirror/synch the 73Gb Drive with the 146 Gb Drive
5. Remove the 73Gb drive and replace the 146 Gb drive
6. Mirror/Synch the x2 146 Gb drives

The rationale is that the software on the 73Gb drive is heavily patched (impossible to clean install on the new server)and we need a simple way to get an exact image to the new rp3440 B

[DCE]

If I understand correctly, you are mirroring from 146 GB drive to a 73 GB drive

Technically, this should work, assuming there is less than 73 GB of data on 146 GB disk, and you mirror everything properly.


If this is the root disk, you will also have to make the disk bootable before you add it to the VG, and run the mkboot commands against it as well.

[Mel Burslan]

This procedure is fine as long as your MAX PE PER PV value is large enough to accomodate the 146G drives and having them fully utilized. Also if the server is not quiesced, i.e., if there are databases and such under heavy use, when you pull out that hot swap disk drive, you are in an indetermined state. My advice would be to shut the server A down prior to taking the drive out.

[Ross McWalter]

Hi there -

Sorry if I wasnt clear - the data is on teh 73 Gb disk. Someone informed me that the remaining 73 Gb on teh larger disk woudl be unusable after the mirror/synch?

[DCE]

I forgot to add

If you are mirroring 73 GB disk to 146 GB disk - this will probably not allow you utilize all of the larger disks capacity unless you were far sighted when you created the vg and made allowances for larger disks then. The parameter is not changable after vg creation and you will have to actually destroy/recreate the vg in order to access all of the disk space

[Mel Burslan]

if you have chosen the default max pe per pv value when creating the volume group on the server A, the one with 73 GB drives, that assumption would be correct, i.e., you can not use the remainder of 146GB disk. It will act like a 73GB disk when you clone them.

[Mel Burslan · Accepted solution]

Also if you happen to read the fine publication by HP, called "when good disks go bad", on page 31, it mentions kind of unsupported utility called vgmodify, available from your friendly hp tech representative, which can change this value.

Anectodal evidence of success rate is nowhere near 100% but may worth a shot.

Hope this helps

[Mel Burslan]

sorry, forgot where the doc. is located on HP's site. So I am attaching it.

[Ross McWalter]

Appreciate the responses - thanks! That being said, the servers are slightly different (server B has more memory and double the # cpus of A). Is there any way to relatively easily get the data to the larger disk?

[James R. Ferguson]

Hi Ross:

The caveats as far as LVM is concerned have been noted by the previous posts.

*However*, if your objective is to replicate the patched envionment and you have identical server architecture, clone your operating system (vg00) with Ignite. Use:

# make_tape_recovery -x inc_entire=vg00 -I -v -a /dev/rmt/0mn

...and perform an Advanced Installation mode install. You can select your target disk and size your vg00 filesystems to your taste, too.

*Before* the Ignite tape is used on the secondary server, 'vgexport' the non-vg00 volume groups with the '-s' switch; and copy the resulting mapfiles to the other server for use after the Ignite tape image is applied.

Following the ignition session, copy the saved mapfiles from the "source" server and use them to 'vgimport' the non-vg00 volume groups.

As for LVM mirroring, it is only and always at the logical volume level.

Regards!

..JRF...

[Mel Burslan]

After having identical vg, lvols and filesystems properly created

host=serverB
for fs in /filesystem1 /filesystem2 #... and so on...
do
cd ${fs}
find . -xdev -depth | cpio -ocx | remsh ${host} "(cd ${fs}; cpio -idcumvx)"
done

goes as fast as your network bandwidth.

another option would be a backup and restore.

[Ross McWalter]

Thanks this should get me going! Appreciate all the help from you gurus!

Cheers!