Adding Filesystem to a Service Guard Cluster

[Mark Treen_1]

Hi Folks

Needing a bit of advice.

One of our operators will be adding a logical volume to one of the volume groups that is part of the cluster configuration. What I need to know (as some of the current posts are either confusing or do not answer the question) are the steps as to how excactly I then incorporate this lvol once created into the cluster config so in the case of a system fallover when the nodes switch this lvol behaves as the others already configured (i.e it switches the secondary node).

As always thanks a million for the effort in responding

Cheers

Mark

[A. Clay Stephenson]

You can add the LVOL and filesystem to the currently active node. Because the metadata are actually recorded in the VGRA('s) of the LVM disk(s), the new LVOL will be available. On each potential adoptive node you will need to add the character and block device nodes under /dev/vgxx. The major/minor device number tuples should match those on the currently active node or at least conform to the same existing patterns.
The better method is to vgexport/vgimport on each node but that is not strictly necessary. Next, you need to modify the .cntl package file on the active node to include the new LVOL and filesystem + mount options. Copy this file to the other adoptive nodes. Make sure that you create mountpoints on each of these nodes for the new filesystem. Don't overlook changes in your backups. Finally, it is a good idea to test failover to each of the nodes.

[Mark Treen_1]

Ok thats a good start thanks so much.

Sorry but there is one thing I do need to add - the new lvol actually has nothing to do with the application in terms of the fact that it will contain application data - the new lvol will act as part of the tivoli solution although in what capacity I will have no involment with.

My brief is to have the lvol active in the cluster config nothing more.

With that in mind would I still need to do anything with the control files (*.cntl)

Mark

[Raj D.]

Hi Mark,

You have to create the Lvol, and FS in adoptive nodes also and also the mount point. Then enter in the /etc/cmcluster/package_name/package_name.cntl file the lvol details, in the adoptive node(s). Check failover.

hth.
Raj.

[Doug O'Leary]

Hey:

To answer your last question first, yes, you'll need to update package control file because, although the new LV isn't used by the application, it must still fail over with the package.

With that answered - here's your process:

1. Create the new LV and filesystem. Mount as appropriate.

2. Update the package control script; ensure you use a unique index number in the arrays.

3. Preview export the affected volume group and and transfer the map file to the adoptive node:

vgexport -p -s -m ${vg}.map ${vg}
scp ${vg}.map ${adopt}:/tmp

4. Note the VG's group file minor number

5. On the adoptive node, export the volume group :

vgexport ${vg}; rm mapfile

6. Create and reimport the volume group definition:

mkdir /dev/${vg}
mknod /dev/${vg}/group c 64 ${minor}
vgimport -s -m /tmp/${vg}.map ${vg}

7. Transfer the package control script to the adoptive node:

(adoptive node)
cd /etc/cmcluster/${pkg}
cp ${pkg}.cntl ${pkg}.cntl.YYMMDD ## or use RCS - much better/cleaner
scp ${primary}:/etc/cmcluster/${pkg}/${pkg}.cntl /etc/cmcluster/${pkg}/${pkg}.cntl

HTH;

Doug O'Leary

[A. Clay Stephenson]

Same answer as before except that it may or may not be necessary to create a filesystem or add the mount and mount options for this filesystem. If you are using a filesystem then my instructions remain the same. Again, strictly speaking a vgexport / vgimport on each node is not necessary though that is the safer method. If there are any special instructions related to Tivoli then these start/stop instructions may well need to go in the packake .cntl file.