Data moving and MC/SG still available , how ?

[Dennis Chao]

Our environment , 2 x N4000 with MC/SG , a FC-60 Disk Array
We want to migrate our data from FC-60 to another storage(IBM Shark). We plan to create exactly the same VG name and LV name on Shark and migrate our SAP data(Use tape backup). When date migration complete , any files or commands should we use to make MC/SG still workable ? Any file should we copy to another N4000(Backup machine)? I am a newly HP user nor to say MC/SG, if some expert can help this , pls. write it more detail and clearly . Thanks for your great help.

[Santosh Nair_1 · Accepted solution]

Dennis,

How are you moving the data..are you creating a new VG, copy over all the data from the old VG and then try to get things working with SG?
If the disks are of the same size, I would suggest that you just add the disk from the new array, i.e the IBM Shark, to the existing VG using vgextend.
Then use the pvmove command to move data from the FC-60 disks to the IBM Shark disks.
Finally, after moving all the data, use the vgreduce command to remove the FC-60 disks from the VG.

If this works out, the only changes that you'd need to make for ServiceGuard would be to change the CLUSTER_LOCK_PV to point to one of the new Shark disks. You'd then also have to re-import the VG on all the other nodes in the cluster and finally, recreate the cluster using the cmcheckconf and cmapplyconf commands.

Hope this helps.

-Santosh

[Sridhar Bhaskarla]

As you are keeping the VG and LV names the same, only thing that you need to change is the cluster lock PV.

Halt the cluster. Do whatever you want. Edit /etc/cmcluster/yourcluser.ascii file, replace the lock device with the new one and do a cmapplyconf. You should be good to go. Remeber to import the volume groups on the both the systems and do a vgchange -c y before you start the packages.

-Sri

[Dennis Chao]

Thanks for your response , but I still have some questions about the meaning of "cluster lock pv" What's the difference between normal disk and the cluster lock pv ? Shall we just edit /etc/cmcluster/yourcluser.ascii file to define our cluster lock pv or we shall use command to write something into the disk(maybe into the superblock or somewhere..)

[Animesh Chakraborty]

Hi,
You don't have to write anything on cluster lock VG.cmapplyconf will take care of that.
steps:
1.halt the cluster
2.Activate the cluster lock vg on the configured node only
vgchange -a y /de/vglock
3.Update the cluster binary configuration file:
cmcheckconf
4.Redistribute the cluster binary configuration file:
cmapplyconf
5.vgchange -a n /dev/vglock
6.cmruncl

Another way:
Do take a vgcfgbackup of /dev/vglock
then
vgcfgrestore -n /dev/vglock /dev/rdsk/c*t*d*

Best of luck
Animesh

[Santosh Nair_1]

Dennis,

The cluster lock PV is a disk that is used when trying to resolve a cluster reformation.
More specifically, if the heartbeat between the two nodes in a cluster stops for some reason, both nodes now think that they are the only survivor.
The cluster lock disk is used to resolve this deadlock. The first node to successfully grab the cluster lock disk reforms is able to reform as a one node cluster.
The other node reboots.

When you define a disk as a cluster lock disk, the VG header is modified to reflect that status.
You really shouldn't have to do anything other than to reform the cluster as Animesh describes above. The proper VG's header will be updated.

Hope this helps.

-Santosh