Need assistance in configuring LVM in HP Service guard Cluster

[rohithroki]

Dear Team,

We have 3 node cluster and in one of the node(DB) we are running out of space and we need to add new lvms for the future data growth.

Kindly let us know the step by step procedure for adding VG/LV in cluster and updating the configurations.

Storage we are using is HP 3 PAR.

Regards,

RKJ

[Matti_Kurkela]

Which HP-UX version (11.31 or earlier)? Which Serviceguard version (run "cmversion" to know)?

Basic Serviceguard or VxVM+Cluster Filesystem? (the latter would allow the filesystem to be mounted on multiple nodes at the same time).

Do you have OnlineJFS available? If you have, it allows extending existing LVs without unmounting them.

Please answer all these questions, otherwise there will be too many alternatives and options to explain in a concise manner.

[rohithroki]

@Mk thanyou for the reply please find the requested details below

HP UX Version : HP-UX 11.31

Service Guard version : A.11.20.00

Filesystem is Vxfs Type SLVM

Online JFS is available .How ever my requirement is to create a new VG and configure lvm and add it to exisiting cluster package.

Thanks in advance for the help !!!!

RKJ

[rohithroki]

Dear Team,

Any help on the above topic is much appreciated

Regards,

RKJ

[Matti_Kurkela]

Step 1: download the "Managing Serviceguard 11.20" book and read it. It is THE most important resource in Serviceguard operations. It has very clear step-by-step instructions for common procedures like this.

http://h20628.www2.hp.com/km-ext/kmcsdirect/emr_na-c03511798-5.pdf

Step 2: present one or more LUNs from your storage, in such a way that all cluster nodes will see them. These will be the PVs of your new LVM volume group. Use the "ioscan -fnNCdisk" command to verify that all nodes see the new LUNs; if necessary, use the "insf" command to create device nodes for them.

Step 3: on one node, create the new volume group. See pages 183-187 of the book. In your case, the storage system will handle any disk mirroring, so you don't have to do it at the cluster level.

Step 4: distribute the volume group information to the other cluster nodes. See pages 185-187 of the book.

Step 5: make sure the new VG is deactivated (vgchange -a n) on all cluster nodes, and switch it into cluster mode (vgchange -c y <name_of_VG>).

Step 6: use "cmgetconf -c <your cluster name> /tmp/cluster.ascii" to dump the current cluster-wide Serviceguard configuration to an ASCII text file. Verify the VOLUME_GROUP parameter: if it already includes your new VG, you can skip the next step. (I think vgchange -c y adds the VG to the binary cluster configuration if cluster already exists and is running, but I might be wrong.)

Step 7: if your new VG is not mentioned in the VOLUME_GROUP parameter in the cluster ASCII file, add it in, then use "cmcheckconf -C /tmp/cluster.ascii" to verify its correctness, then run "cmapplyconf -C /tmp/cluster.ascii".

Now the volume group is ready to be added into the configuration of your cluster service package.

To add it, you'll first need to determine the type of the package: is it a legacy package or a modular package?

Modular packages are described starting from page 232 of the book, and legacy packages starting from page 307.

For legacy packages, you must halt the package before you can add a new VG to it. For modular packages, it is possible to add a new VG to a package while the package is running, but you must be very very careful: if there is a configuration error that will prevent Serviceguard from bringing up the new VG and any filesystems added to the package configuration, the entire package will fail when you apply the configuration change. See page 315-317 of the book for a general description of package reconfiguration, and Table 16 starting on page 320 for the kinds of reconfiguration operations you can do for the packages and the requirements for those operations.

For modular packages, the procedure is fairly simple: use "cmgetconf -P <package name> /tmp/package.ascii" to dump the current binary package configuration into a text file, then add the necessary vg and filesystem parameters to the file, use "cmcheckconf -v -P /tmp/package.ascii" to verify the correctness of the updated text file, then use "cmapplyconf -v -P /tmp/package.ascii" to apply the new configuration.

For a legacy package, you don't actually need to edit the package configuration file; the VGs and filesystems and their mount locations are defined in a package control script instead. You must halt the package, locate the control script, add in the VG and filesystem parameters for your new VG and its filesystem(s), copy the updated control script to all the cluster nodes, and then restart the package.

In older Serviceguard versions and most examples,, the name of the control script for a legacy package used to be /etc/cmcluster/<package name>/<package name>.cntl, but it is now technically possible to use other names for it. If necessary, you can use "cmgetconf -P <package name> /tmp/package.ascii" to dump out the package configuration in text form, and find the pathname of the control script in there.

(So far, this forum has failed twice when I've attempted to post this, and lost the text I had written. Let's see if third time's the charm...)

[rohithroki]

@Matti_Kurkela

Thank you for the detailed steps .Will go through the same and update you if in case i have any futher questions/doubts.

Regards,

RKJ