some questions about data migration with vxvm

[franky_sun]

Hi experts there,

I have two AIX 5306 hosts using vxvm +vcs +vvr to run Oracle.  Now we want to migrate all data to a new array.  The size of source disks are not equal - from 50G to 300G and the new array must set all luns 144G.  All the volumes are concatenated.  So I have got some questons as below,  Would you please kindly help me??

1, I want to use vxvm mirror to migrate data,  which way is the best practice?  vxmirror? vxdg mirror? or vxassit mirror?

2, How to deal with the vcs and vvr?

Further more, If the volumes are complex (e.g., VxVM RAID-0, RAID-5, etc.) which command can use?

Best wishes and Many thanks!!!

[franky_sun]

Can I use " vxmirror -g mydg -a newdiskmedia1 newdiskmedia2 ... newdiskmedian" to migrate all my volumes to the newdisks?  Could it be right?

[shellhead]

I've performed quite a few data migrations using Veritas Volume Manager. Please read my article for SysAdmin Magazine, which has been archived here:

http://www.theillien.com/Sys_Admin_v12/html/v13/i08/a1.htm

If you have any questions, or require some consulting services, please let me know.

[franky_sun]

hi shellhead,

Thank you for your reply but I can not open your link.  Do you mean the article as below?

http://www.samag.com/documents/s=9364/sam0408a/0408a.htm

I have read it and found some parts not so clearly， especially how to migrate data.  In my environment， the capacity of "new_vm_disk" and "old_vm_disk" are not equal！  So I wonder if I could use the “vxmirror” command as I guessed.

[shellhead]

Yes, that's the correct link to the article. Mirroring the volumes and then breaking the mirrors is one method for migrating data.

You can only use the vxmirror command to mirror concatenated volumes -- unless newer versions of VxVM offer this feature (I don't have access to version 5.x).

You can only mirror from a smaller LUN to a larger LUN, but not the other way around. You can always add additional disks to a diskgroup, and vxmirror will automatically mirror to the next available disk. If you have some space leftover, you'll have no choice but to either live with the wasted overhead, or put it to some use.

If you need to mirror a complex volume, such as a striped volume, you'll need to manually create a concatenated plex and attach it the volume to create a mirror.

[franky_sun · Accepted solution]

I tested the migration steps and found that "vxmirror" can only copy simple volumes that had got one plex and one sd.  So I can only use "vxassist" command to do my work.

[Keeg]

"You can only mirror from a smaller LUN to a larger LUN, but not the other way around."

That's not true. It my be a limitation of vxmirror but you can do it with vxassist. I just finished a large data migration and this scenario came up several times.

We are using VM 4.1 on Solaris 9 & 10

These were the commands I used to create the mirrors:

vxassist -b -g <disk group> mirror <volume name> layout=nostripe

To specify which array to mirror to use:

vxassist -b -g <disk group> mirror <volume name> layout=nostripe alloc=enclr:<enclosure name>

To specify which disks/luns to use:

vxassist -b -g <disk group> mirror <volume name> layout=nostripe alloc=disk_1,disk_2,disk_3

To specify which disks/luns NOT to use:

 vxassist -b -g <disk group> mirror <volume name> layout=nostripe alloc="\!disk_1,\!disk_2,\!disk_3"

You can use vxassist to break a mirror but it will leave the subdisks and plexes. This command breaks the mirror and removes the plex and subdisks:

vxplex -g <disk group> -o rm dis <plex name>

Hint: Use identifying names when adding disks/luns to a disk group that will be migration targets. If you let Veritas assign default names it's harder identifying which mirror you want to get rid of. By default subdisks get their name from the source disk which get their name from the disk group. Using identifying names makes vxprint much easier to read and old plexes easier to identify in vxprint.

Hope this helps.

Don

[Eric_Gao]

not quite familar with vxmirror, but I always do this kind of migration step by step.

1.  add new disks (from new array)

2.  create subdisk and plex with identical size

3.  attach the plex to volume

4.  after fully mirrored, split the old plex, destroy it and then old disks can be taken out of vxvm.

5.  grow the volume to fully utilize the capacity of new disks.

When you have the volume that consists of many subdisks, across many disks, above steps are tidious and boring I know, but it works and also let you know more what vxmirror is working.

Message Edited by Eric Gao on 02-24-2009 08:35 PM

[Xentar]

Hi all,

After read the link from shellhead, there is prerequisites about the file system for database should be raw device, otherwise, it's not possible to migrate with VxVM mirror migration. I would like to ask that why data file with file system is not support with VxVM migration?

Or if the database is in offline status, can we use this method?

Regards,
Xentar