EMC BCV Restore (Question)

[Dario_1]

Hi Everyone,

Does anyone have the steps required to restore from a bcv to the standard device for an EMC Symmetrix connected to a HP server running HP-UX 11.00 and Symcli 5.1.1. We are totally new to EMC. The implementation was done by the EMC people. In order to learn how to restore, we created a test DB (oracle) that the system is backing up every day. This morning I stopped the test DB, umount the file system, deactivate the volume group, executed an incremental establish (symmir -g groupname est -nop)and a cold split (symmir -g groupname split -ins -nop). Activated VG, mounted FS and deleted the information under the FS. That part was fine.

Now, I am trying to restore from the BCV.

1. umounted FS
2. deactivated VG
3. Restore (symmir -g groupname -full restore -nop)
4. monitored the restore using (symld -g groupname list)
5. Split again (symmir -g groupname split -nop)
6. Activated and mounted

But I did not see the data. I also tried with the FS mounted and VG activated and I was able to see the oradata directory but when the restore was done, the FS was empty.

I am restoring in the backup server. Am I doing it correctly? Do I have any step missing in my instructions?

Thank you for your help.

Regards,

Dario



Regards,

[Michael Tully · Accepted solution]

Hi Dario,

When you refresh from BCV and mount onto the server these are the procedures we use. I am assuming your using 'srdf' and have R1/R2/BCV volumes.

shutdown databases
umount all filesystems
vgchange -a n /dev/vg?? (all volume groups being used)
vgexport /dev/vg?? (all volume groups being used)

On the system controlling your EMC
symmir -g query (should be split)
symmir -g establish (to get fresh prod copy)
symmir -g query -i 10 (check progress)
on completion
symmir -g split
vgchgid (all LUNS being used. You can place all LUNS into a files and cat the the in)
vgchgid `cat myrawdiskfile` (for each vol grp)
mkdir /dev/vg??
mknod /dev/vg??/group c 64 0x040000
vgimport -v -m myvg.mapfile -p `cat mydiskfile`
if successful, drop the -p
Activate the vol grp
vgchange -a y /dev/vg??
You need to run an 'fsck' on all logical volumes you intend to mount.
mount the filesystems (sizes should be okay)
Your database should start

HTH
Michael

[Massimo Bianchi]

Hi Dario,
your procedure seems ok, but i have some notes.


bcv operation occurs at disk level, so you must ensure that target disk are NEVER mounted.


So, when you do the backup, umount BCV pairs and deactivate the vg.


When you do the restore, deactivate the original disks (source and BCV). This is necessary because during BCV operations BCV disks are not seen by any hosts, and original disks are being rewritten.



First: check that the backup part went fine: after a cold split, mount the FS on the BCV to test and see if data is there. If it is a cold split, you have not to use fsck, keep in mind. If you have to use it, than we are missing something.

For the restore part.... bho! ask EMC, they should have released a procedure with you.

Your steps looks OK, i do not see why your procedure do not work. Try the other checks before the restore...

Massimo

[Dario_1]

Michael and Massimo:

Thank you for the answers, I will be running the test this afternoon again and will assign points as well. I will let you know the result.

Thank you again.

Regards,

Dario

[Dario_1]

Michael and Massimo:

Thank you for your help. Finally, I was able to come back and finish this. My problem was that I was only deactivating the volume group in the target server and leaving the backup server active. I deactivated in both and worked just fine.

Thank you again.

Regards,

Dario