Host based SAN migration

[Waqar Razi]

We have an MC Service Guard CLuster running on two Red Hat Linux 5.1 nodes. We are migrating SAN and this will be a host based migration. The SAN Administrator have presented new LUNS and we have to migrate the application file system /application to it.

Can some one please give me the best approach to do this? Please keep in mind it is Service Guard Cluster.

Thanks for your help.

[Steven E. Protter]

Shalom,

Several good approaches.

1) Bring the applications down, and use SAN based tools to replicate the data.

2) Bring the applications down and use tar or scp -rp to copy the data system to system.

3) Present the new LUNS to the system, replicate the data, bring the applications down and umount the filesystems,then change the /etc/fstab entries, and mount the filesystems and bring up the applications.

SEP

[Hemmetter]

Hi

do you use LVM?
rescan all scsi buses by
# echo "- - - " > /sys/class/scsi_host/host?/scan.

Then add your new LUNs to VG (pvcreate/vgextend) and use pvmove(1) to migrate your data online to the new LUNs.
After that vgreduce if needed. Keep in mind to tell the cluster about the new cluster-lock-disk if it is subject to migration as well.

rgds
HGH

[Waqar Razi]

Thanks for your reply. Can you please provide me more details on the pvmove? If possible, step by step direction would be very helpful.

[Hemmetter]

First of all do you you use LVM on your cluster? How familiar are you with LVM?

Can you give some more detail of your current and you future setup?


rgds
HGH

[Waqar Razi]

I am familiar with LVM. I have bacground with HPUX. In HPUX, we can use mirrors to mirror the data and then break the mirror.

I want some thing equivalent in Linux. I tried lvextend -m 1 but it doesnt work.

[Kapil Jha]

I would say best id to use SAN tools they would be having something which would replicate the data (make sure your application is down).

If you want to use LVM its fine, and when u say lvextend is not working what is the error you getting.
have you added the new disks in VG properly?

BR,
Kapil+

[Hemmetter]

Hi again,

what you tried with lvextend -m 1 to add a mirror to lvol is exactly that what pvmove does internally. It creates a mirror and when it is synchronized it will reduce it by the old pv.


# pvmove [-n/--name LogicalVolume] [SourcePhysicalVolume[:PE[-PE]...] [DestinationPhysicalVolume[:PE[-PE]...]...]]


You can pvmove either all physical extends off a pv to a new pv or you can do it per lvol.

Have a look at pvmove(1).

rgds
HGH