Reinstalling HPUX without changing LVM configuration.

[Satish K. Vanimisetti]

Hi,

I am getting ready to re-install an HP workstation running HPUX 11.11, and want to gather some information before I take the plunge. (I mostly administer Linux/FreeBSD machines, and have limited HPUX experience).

The home partition on this workstation was moved from local disks to a fiber channel RAID5 array. The array is administered by someone else, and shows up as three disk devices on the workstation. These devices are configured into a volume group, and a single (1.86TB) logical volume (which uses all available space on the array) is configured in the volume group. We have to use LVM for the fiber channel array so we can use multipathing.

Now, I want to repartition the local disks with a better layout. (More swap, bigger partition for software, no home partition).

I want to know:

1. If there is a way to save and export the LVM configuration for the fiber-channel volume group? Can I import this configuration after I have re-installed the OS, and mount the logical volume without having to re-format and restore data?

2. If the 2*RAM rule for swap space applies in HPUX. (The workstation has 16G of RAM and is used for CFD simulations).

3. If anyone that has attempted something similar has any tips/tricks/gotchas.

Thanks a bunch in advance.
--
सतीश् (Satish).

PS: My apologies if this description is even more vague than I believe it is. Please let me know what details will be helpful, and I will post them.

[Pete Randall]

You can use vgexport and vgimport to save and then reconnect the VG containing /home. I would suggest:

vgexport -s -m /tmp/vg01.map /dev/vg01

-rebuild vg00-

mkdir /dev/vg01
mknod /dev/vg01/group c 64 0x010000
vgimport -s -m /tmp/vg01.map /dev/vg01
vgchange -a y /dev/vg01


As far as re-installing just to change your LVM structure, it seems to me that you would be better served to download and install Ignite, then run make_tape_recovery to create a bootable backup of vg00. When you boot off the tape (which has to be local, by the way), you can interact during the restore process to re-size your logical volumes.

Finally, with large memory systems, nowdays, the 2*RAM rule for swap space is not strictly observed. Many times you can get away with 1*RAM or even less. Check what you have currently.


Pete

[asif munir]

Hi,

If you have external storage, then what you can do is make map files for your volume groups, i.e vgexport -m vg01.mymap /dev/vg01, if you are going to reinstall then move the mapfile of to another machine. After you have re-installed the os and re-present the same luns from the external storage to the os you do a vgimport -v -m /vgsan01.mymap /dev/vg01 /dev/dsk/cxtxdx ....
After this all you do is vgchange -a y /dev/vg01

After you re-present your external luns back to the server make sure you DO NOT do a pvcreate on the disk.

With regards to SWAP due to the large amount of physical RAM you have then you may not need to follow the 2xRAM rule. But again this entirely depends on the applicaiton that will run on the machine and how memory intensive it is.

Asif

[James R. Ferguson]

Hi:

You should 'vgexport' and 'vgimport' your non-vg00 volume groups. In this fashion, your data will be preserved. Have a look at the respective manpages. You can use the '-s' and '-m' options during the export and import process to include the VGID in the mapfile. In this way, your import will discover the correct physical volumes to be associated with a particular volume group.

If you have alternate physical paths, remember to use 'vgreduce' and 'vgextend' to promote the appropriate path to be the primary one so that high-availabilty via multiple controllers is achieved.

As for swap, the 2X rule is not fixed. Enable pseudoswap by configuring the kernel parameter 'swapmem_on=1'. This counts 75% of your physical memory for process swap reservation accounting.

If you have a small primary device swap (required) then you can add additional swap on another non-vg00 device to satisfy swap needs.

Regards!

...JRF...