partition a disk NOT using LVM !!!

[Jose_11]

Hi Boys and Girls,

How can I create different slices in a disk without going thru LVM?
I mean that I want to partitioned one disk to two or more parts, and make file system on each part, so I can mount each file system to a different mount point ?

Any help is greatly appreciated.

[Sridhar Bhaskarla · Accepted solution]

Hi,

Not it is not possible with HP-UX later to 9.04.

Either you need to have LVM or VxVM to do it for you.

-Sri

[Stacey Rippetoe]

Hi Joseph,
Not with HP-UX (other than VxVM). That's what lvm is all about. I there a reason you don't want to use lvm?


S.

[S.K. Chan]

Take a look at ..
# man newfs_hfs
You can create a whole disk partition (non-LVM) but you can't slice it up.
# newfs -F hfs /dev/rdsk/cXtYdZ
# mount /dev/dsk/cXtYdZ /testdir

[Arockia Jegan]

Through sam,

sam---> Disks and File systems--> Disk Devices---> Actions ---> configure---> Not using LVM

[Arockia Jegan]

Joseph,

You can have only one partition per disk if you are not using LVM/VXVM and using the whole disk approach.

So you can have only one file system per disk. It's the limitation of this whole disk approach.

Create the file system by using,

#newfs -F hfs /dev/rdsk/cXtXdX

[Jose_11]

Hi All,

Thanks for prompt reply .

Stacey, the reason I want to use hfs file system is performace reasons.

Thanks anyway you all,

Joseph

[A. Clay Stephenson]

If you want to use hfs it must be because you want to reduce filesystem throughput. Vxfs has outperformed hfs now for quite a long time. You would also find that LVM introduces only a small bit of overhead that is almost always more than made up for with the ability to stripe.

The only overhead associated with LVM is one extra redirection. Whenever an I/O request is made to an LVM block a kernel map is consulted to say that this is really block nnn of /dev/dsk/c1t2d3 and then the I/O requested is passed off to the actual disk device I/O function. Because the redirection is but a small fracxtion of the total I/O request the overhead is actually difficult to measure in real life especially is you are striping the LVOL over multiple paths and devices.

[Bill Hassell]

As Clay mentioned, LVM performance (and specifically HFS) are VERY different that they were early in 10.01 and 10.10. There are only three methods to manage disk sectioning in 11.xx today: whole disk (which is impractical for the boot disk), LVM (which is very fast, and virtually indistinguishable from whole disk performance) and VxVM which is the Veritas Volume Manager, supplied with 11i but available from Veritas for 11.0.

Many, many years ago, there were sysadmin stories that LVM extracted a performance penalty and indeed, without the required patches, that was true, especially on the very slow machines from 10 years ago.

As far as filesystems installed on the partitions, HFS was faster than VxFS, again, many years ago, but is not true today. The VxFS filesystem is stable, fast and recovers from an abnormal reboot in seconds rather than hours (with HFS).

[Trond Haugen]

Just to add to Bill's story about the old times. There where then posible to partition a disk thru the disktab entry for servers. But those where fixed. On the workstations you could tweak SDS (software disk striping) to acomplish it. So LVM was a verry welcome ting.

Regards,
Trond