MC Service Guard & LVM

[Indrajit Bhagat]

suppose there are two VG vg01 & vg02 that are used by cluster. now suppose i want to add two new physical volume to vg02 which is used by application package.

How can i proceed?
whether i have to stop the cluster or can i add it online.

can i have a steps for this.

[Hasan Atasoy · Accepted solution]

hi ;

1. assign lun for both system.
2. vgextend vg02
3. lvextend or lvcreate
4. if you create a new fs by lvextend change cluster configuration file on both node. ( /etc/cmcluster/pkgname/*.sh )
5. mount your logical volume.

6. Update the LVM information for the vg on all other nodes in the cluster:

# vgexport -p -v -s -m /tmp/vg??.map vg??
# rcp -p /tmp/vg??.map nodeB:/tmp

On the other node(s):

(first, note the minor number of the vg group file, 0x??0000, as you will need it)
# ll /dev/vg??/group
# vgexport vg??
# mkdir /dev/vg??
# mknod /dev/vg??/group c 64 0x??0000 (same minor # as before)
# vgimport -v -s -m /tmp/vg??.map vg??

[Steven E. Protter]

Shalom,

Network issues prevented me from resopnsing earlier.

The general model here is the package will do the following:

1) Activate the volume group.
2) Mount the filesystems if any on the logical volumes within the volume group.

So you are saying you need additional data area for a new package wthat will not perform item 1 above.

node1
pvcreate the new disk
vgextend vg02 to include the new disk in the volume group.

lvcreate/lvextend any logical volumes needed on the disk.
newfs any filsystems needed on the disk.

Do a vgexport -p (preview) mode to get a map file that you will vgimport to the second node.

Copy the map file to node2

vgimport using the map file and all configuration done on node1 will be available on node2.


SEP

[Asif Sharif]

Hi Bhagat,

LUN Assignment to nodea and nodeb
# ioscan â fnC disk
(execute on both nodea and nodeb)
Insf -e
(execute on both nodea and nodeb)
Inititialise LUNs
# pvcreate /dev/rdsk/cXtYdZ
(Perform for all new LUNs)
Extend LUNs into the VGs
# vgextend /dev/vgname /dev/dsk/cXtYdZ
Extend the Logical Volume size
# lvextend â l /dev/vgname/lvol1

Extend the file system size according to the new Logical Volume size
# fsadm â F vxfs â b m /mountpoint

On nodea, create the VG map files

# mkdir /tmp/mapfile
# vgexport â p â v â s â m /tmp/mapfile/vgname.map /dev/vgname


FTP map file in /tmp/mapfile to nodeb

On nodeb, export those VGs that require file systems extension
# vgexport /dev/vgname

On nodeb, Create VG directory and VG device file
# mkdir /dev/vgname
# mknod /dev/vg10/group c 64 0x0a0000

On nodeb, Import VGs
# vgimport â v â m /tmp/mapfile/vgname.map /dev/vgname

For more detail also see:

http://docs.hp.com/en/B3936-90079/index.html

Regards,
Asif Sharif

[Indrajit Bhagat]

Thanks for replying.

I wants to know, while adding a disk to volume group which has been used by cluster.
at that moment we have to stop cluster to while online only we can add the disk and increase the file system size.

Also what is an exclusive mode in mc service guard cluster.

[Hasan Atasoy]

hi indrajit ;

if you are running an active/passive package ( not a rac ) you dont have to stop the cluster or package in order to add disk to volume group . add disk in active node and extend your fs ( if you have online jfs license ) or create a new fs and mount it. and update passive node by export import process.

but if you are running an a/a cluster package you need to stop at least one of your nodes to add disk to vg but it requires a OLR patch. if you dont have that patch , in order to add a disk to a vg you need to stop the package.


from the man vgchange manual

"vgchange -a y " Activate each specified volume group and all associated physical and logical volumes for read-write access. If a volume group is marked as part of a high availability cluster, it is activated in exclusive read-write mode, as for the -a e option.


"vgchange -a e"

Activate each specified volume group and all associated physical and logical volumes for exclusive read-write access. The volume group must be marked as part of a high availability cluster, and the availability software must be running on the system; otherwise, the volume group is not activated.

so if you are running a vg ınder control a cluster, activating that vg is done by -e


Hasan