Expand disk in cluster configuration

[Dmitry Skutin]

Hello,

I have two HP-UX 11.23 boxes working in cluster configuration and EVA5000 as storage array. Now I have a task to expand one disk. On EVA side it's no problem.
But what I should to do on OS side if I use LVM but not VxFS?

Thanks.

[R.K. #]

Hi Dmitry,

Check out this link:

http://deepees.wordpress.com/2009/02/18/hp-ux-adding-a-disk-to-an-mc-serviceguard-cluster-volumegroup/


May be you will get what you are looking for.

Regards.

[Matti_Kurkela]

LVM but not VxFS?

So you have an application that uses /dev/vg/[r]lvol* directly?

In that case, first make sure that the application can use the expanded disk space. Often you may need to do something to tell it to use the new space. Find out what is the application-specific procedure and if there are any limitations to it.

You can then present new LUN(s) from EVA to HP-UX, pvcreate them, and add them to the VG as normal. Then use lvextend to increase the size of the application LV as needed, but do NOT use fsadm on the application LV: do the application-specific procedure instead.

Increasing the size of existing LUNs is not recommended: it's much more difficult to make HP-UX recognize the new size of existing LUNs than to make it accept new LUNs altogether. When a new LUN is added to the VG and then used to expand the respective LV(s), the final result is the same: the LV grows in size.

The whole point of LVM is to prevent the filesystems and applications from being limited by the actual physical size of your storage unit(s).

MK

[Dmitry Skutin]

As I see it's actually the same as add new disk to volume group: stop application which used the disk, get exclusive access to volume group for reconfiguration and so on.

Is there another way without these steps i.e expand disk without stopping application?

[R.K. #]

Hi..

If online JFS is installed, then probably you do not need to stop packages.

# swlist | grep -i online

Sam might not allow you, but command line will.

[Matti_Kurkela]

If VxFS is not used at all, OnlineJFS is not applicable to the situation.

It would be much simpler to answer to this question if you could tell us the name of the application and what you actually *are* using on the disks. Is it raw application data on the LVs, a HFS filesystem, or something else?

You probably can run lvextend without stopping the application. But you must read the documentation of the application to know if it can deal with disks suddenly becoming bigger while the application is running, or not.

For example, if you're running an Oracle database with raw data LVs, some Oracle DBA could probably tell you the necessary SQLplus incantations to make the database engine use the enlarged LV while the database is running.

MK

[Dmitry Skutin]

Matti,

I have Oracle database with raw data on LVs.
If I correctly understand you I should do the following:

1. Expand disk on EVA
2. With lvextend command expand disk
3. Resize Oracle's datafiles

Is it enough?

Thanks.

[Vishu · Accepted solution]

Hi Dmitry,

yes, you are right. But make sure you have to expand it on both the nodes in the cluster, if the LV belongs to the shared VG. you can perform this way:-

1. Expand disk on EVA.

On first node,
2. do, lvextend followed by fsadm/extendfs on the LV. check with lvdisplay.
3. create a map file for that VG in the preview mode.
4. rcp that map file to the other server.

On second node,
5. note down the minor no. of that VG.
6. vgexport that VG.
7. vgimport that VG with the map file copied from the first node (use -s option also).
8. do lvdisplay to check the LV on this node too.

Now, you can resize oracle datafiles.

Thanks

[Vishu]

Hi again,

you have not assigned points to the reply of any of your question. So, please spare sometime to assign points to them.
This is a good way to say thanks.