disk removing

[Murat SULUHAN]

Hi all

I have 2 HP 2430 systems with mirrored 2 DS2300 disk arrays(4*36 GB per array). These systems run with MC/ServiceGuard but our clustered package needs very small hard disk space.

So I want to remove 3 disks from each array to use with another system.

But I cannot find out the remove procedure about cluster and mirrored disk, help me please.

Best Regards
Murat SULUHAN

[Slawomir Gora · Accepted solution]

Hi,

You have to check which volume grup belongs to cluster and distribution of your logical volumens. After that you must:
- shutdown package(s).
- activate group on node1
- remove wolumens - if needed.
- reduce vg group (vgreduce) on node1 - remove disks
- export vg configuration in preview mode
- copy mapfile to node2
- remove vg config (vgexport)
- create new definition for vg (mkdir /dev/vg.., mknod)
- import vg - vgimport
- backup vg configuration vgcfgbackup
- deactivate group
- start package

2. Maby simpler will be:
- backup data
- remove vg
- create new vg with only 1 disk per DS2300
- import/export ...

[Franky_1]

Hi,

first of all make a backup :-)

Then you can remove the old volume group and create a new one with just one disk

Regards

Franky

[Sridhar Bhaskarla]

Hi Murat,

Basically the procedure is the same as other except that you will have to make changes on the secondary node also. If the filesystems are spread all over the disks, then move them to fewer disks and free up the other disks. There are few scenarios on how you can move the data.

1. PV1 has LV1 and PV2 has LV2. PV1 has enough space to hold LV2. In this case you would do

lvextend -m 1 LV1 PV1
lvreduce -m 0 LV1 PV2
vgreduce vgxx PV2

2. LV1 is spread across PV1 and PV2 so as LV2. But PV1 has space to hold both LV1 and LV2. In this case you cannot use mirror/reduce method. Use 'pvmove'

pvmove PV2 PV1
vgreduce vgxx PV2

Look at 'pvmove' for more information.

The above is online. You don't have to bring down the packages. But if you are using 'pvmove' make sure you do it during the maintenance window as there is little risk associated with it.

Once the vgreduce (ensure you are not reducing the cluster lock disk) is done, then create the map files and re-import the VG on the other node.

#vgexport -p -v -s -m /tmp/vgxx.map vgxx

Copy vgxx.map to the secondary node.

On the secondary node:

#vgexport vgxx
#mkdir /dev/vgxx
#mknod /dev/vgxx/group c 64 0x0?0000 (get the minor from the primary node)
#vgimport -v -s -m /tmp/vgxx.map vgxx

When you get a chance, do the failover testing to make sure it works.

-Sri

[Murat SULUHAN]

Hi all

Sorry for delayed update


Thanx for each response, I can successfully removed the disks.

Thanx for everything

Best Regards
Murat

[Sridhar Bhaskarla]

Murat,

Interesting enough.. You did not need to shutdown the packages or the cluster unless you are reconfiguring the cluster lock disks.

-Sri