What to do on Serviceguard when migrating to a new storage

[f. halili]

Hello everyone,

We have one cluster comprised of two nodes and they have one package. We are moving them to a new storage one server at a time. I have the ff:

1) Created new VG's, LV's, FS'es on active node.
2) I plan to umount the existing FS w/c are part of the package, mount the NEW VGs/LVs to the existing mount point and restore the data.
3) Now I'm lost on what to do on the ServiceGuard areas before and after the restore.

Please provide most detailed steps as possible.


THANKS,
f. halili

[Mel Burslan]

In your package control files, you have such statements:

LV[0]="/dev/vgdb1/lvol1" ;FS[0]="/oracle" ;FS_MOUNT_OPT[0]="-o delaylog"

if the LV portion of this is not changing, you do not need to do anything with the package control files.

If you are using lock disks for cluster quorum, make sure the vg for the lock disk has the same name, otherwise you need to change this in your cluster config file and recompile the cluster binaries by cmapplyconf command.

also make sure your new volumes are marked as cluster aware volumes.

vgchange -c y vgname

for all cluster controlled volume groups.

[Mohanasundaram_1 · Accepted solution]

Hi,

You should check the "using serviceguard" manual.

Basically you need downtime for including the new storage in the cluster configuration.

1) Create VG LV and fs.
2) Import the VG to the other node.
3) make copy of cluster ascii file and make the necessary changes in the copy.
4) When you get the downtime, perform cmcheckconf using the new modified config file.
5) If the lock disks are also changing, then make the changes accordingly before cmcheckconf
6) then perform the cmapplyconf

With regards,
Mohan.

[Devesh Pant_1]

Halili,

You will have to come up with a better plan than suggested above unless you do not care about the data in the existing storage.

I would say instead of doing one server at a time you definitely have to present the storage to both the servers and then migrate the volumegroups to the new LUNS
1. Using the pvcreate on the new luns then 2. Using vgextend to the new LUN
3. Do the vgreduce for the old storage LUN

All these can be done for each of the volumegroups.

I would expect you to take this up while the cluster is down.

thanks
DP

[f. halili]

thanks for all the inputs.