HP-UX LVM to Recognize EMC Meta LUN

[Lynn Hsu_2]

Hello,
I used EMC's meta LUN to increase a LUN size on EMC's CX500 storage. HPUX diskinfo recognized new size. vgdisplay did not show the new size. I was told by both EMC and HP that I have to redo "pvcreate" on the device to make it visible. This means I have to back up the data, then do the LVM stuff, then restore data... too much down time for a production LUN.

As a workaround, I created a new LUN then added to the vol group, then extended the LV. This works fine.

Wondering what is the use of Meta LUN if one has to distory data to reconfigure LVM? Any suggestion?

Thanks in advance,

Lynn

[Bill Costigan · Accepted solution]

With LVM the size of the PV is fixed at the time it is created. Moreover, the largest size a PV can ever be is fixed at the time the Volume Group is created.

(extent size X max extents/PV)

Therefore LUN expansion is useless with LVM. However, you could create a VxFS file system directly on the raw EMC LUN (not using LVM at all) and then if the LUN is expanded, JFS can be used to extend the filesystem.

I beleive that the veritas volume manger does support the expanding of LUNs as does windows and other file systems.

[sathish kannan]

Hi Lynn,

Bill is correct. Disk size is set while creating the disk using pvcreate and max PE per PV parameter in VG.

I would recommend to add a new LUN to LVM rather than extending it (META lun) in storage array.

Regards
Sathish

[Matti_Kurkela]

About the usefulness of Meta LUN:
you can have many servers with different needs sharing one big centralized storage system.

For example, we have some ServiceGuard clusters which essentially need shared storage for application binaries and basic configuration data only: everything else is in a database on another host/cluster. So we need some very small FibreChannel-connected storage units.

On the other hand, we have some databases in a terabyte-size class.

I'm not a storage system administrator, but in my understanding it's like this: when designing a configuration for EMC Symmetrix storage systems, one of the first things is to pick a size for a basic LUN. This size is apparently not too easy to change without first moving all data out of the storage system.

It seems to be easy to assign EMC storage space in multiples of basic LUN size, but assigning a space _smaller_ than one basic LUN is more difficult (or may be impossible).

On the other hand, if the basic LUN size is something around 10 GB, a terabyte-sized database would need _one hundred_ LUNs. This quickly becomes a maintenance nightmare.
Fortunately, the basic LUNs can be grouped into Meta LUNs, so instead of a hundred basic LUNs, I can ask our storage admin to assign us a terabyte in one big chunk, or maybe as a ten 100GB Meta LUNs if we expect that some of that capacity requirement is only temporary.

When creating LVM volume groups for use with a big centralized storage, it is essential to think carefully about the MaxPE and MaxPV parameters, so the VG can accept larger LUNs than initially used if necessary. On the other hand, over-large MaxPE increases the size of LVM configuration backups in /etc/lvmconf, which may cause problems if your root filesystem is small.