disk path change

[kholikt]

Hi,

I have a two nodes linux cluster running on MC/SG. It is using volume group. What is the proper step if I want to move the MC/SG to a different set of disks (path will change). Do I still need to do vgexport of volume group to map file. Should I reapply the clusterconf

[Matti_Kurkela]

Linux LVM is different from HP-UX LVM: in Linux, map files don't exist and "vgexport" means a very different thing.

You can move the data with four steps:

1.) add the new disks as new PVs to the VG (first pvcreate, then vgextend)
2.) use pvmove to move the data to the new disks
3.) once the old disks are free, use vgreduce to remove them from the VG
4.) on the inactive node, run "vgscan -vv" make it learn the new VG configuration and to verify that it can see the new disks.

If your package configuration includes direct references to the old disks (e.g. disk availability monitoring), you must change these to point to the new disks instead. But if your package configuration only references to the disks using the VG name, there is no reason to modify the package configuration at all.

This is because Serviceguard for Linux does not have a kernel-based "cluster mode" for VGs like HP-UX Serviceguard has.

NOTE:
Linux LVM tools *do* include a "vgchange -c y" function, but it does *not* work with ServiceGuard.

It is for RedHat Cluster only: if you switch a VG to cluster mode and don't have RedHat cluster daemons running, you'll just make the VG inaccessible. Undoing this mistake non-destructively requires a little trick that is documented in the RedHat Knowledge Base:
http://kbase.redhat.com/faq/docs/DOC-3619
(Guess why I know this... :)

MK

[Prasanth V Aravind]

another way

1. add new disks
2. create vgs & fs
3. mount these file systems with new mount points
4. shutdown pkg & mount old filesystems to a temp mount point, so that data won't get changed.
5. copy data from old to new with dd or cpio
6. unmount all filesysems & put new entries in pkg cntl file.
7. start package.


Gudluck
Prasanth

[kholikt]

Hi,

The migration method will be using EMC sancopy so basically it is just block level copy of the LUN from old storage to new storage. Once the sancopy complete, the LUN will have exact image of the source.

The problem is in the cluster.conf it point to the /dev/sdb1 for the lock lun. The lock lun will be using the sancopy method. I would like to know how to update the config file after we complete the migration.