On Mon, 2002-04-08 at 16:16, Ryland Bingham wrote:
> Is this limit for LVM in Linux in General?
>
> >>There is a fdisk table limit is 4 primaries, one of which may
> >>be of extended type, and a total maximum of 13 to 16 depending
> >>on hard drive type (IDE v SCSI).
> >>
> >>-- Russ Herrold
>
> Using redhat 7.1 and 7.2, I've created as many as 37 partitions on a
> single drive. I think the theoretical limit is 64. I install mainframe
> OS's (z/OS z/VM VSE) in Linux using Flexes emulation. To achieve the
> best I/O performance on the emulated DASD I need many raw(no linux file
> system) partitions(up to 30 on a 80GB drive) to match the size of a real
> Mainframe Dasd devices.
Ah.Linux S/390 in zVM. I've not really tried carving up a minidisk with
partitions (in fact, the most real hands-on experience I've had with
Linux S/390 is with Hercules emulation).
> I'm very interested in LVM if it can be used to dynamically handle a
> large number of partitions. I've not found any practical guideline on
> implementation only white papers and the such. Any body got experience
> or references?
Yes, LVM can be made to create virtual growable/shrinkable partitions
across many defined block devices. LVM is an extent-based virtual
partitioning layer that treats multiple "physical" devices as a single
"volume group" from which "logical volumes" may be carved out.
LVM volumes may be striped across all available physical disks in a
volume group if need be for performance reasons. Multiple volume groups
can be used to ensure volume<>spindle separation.
LVM volumes are "raw. They appear as /dev/{volumegroup}/{lvname}, and
can be used as raw devices for database extents.
You know as well as I how convoluted device contention can become with
virtual volumes. It's rather difficult to keep track of what spindles
have which pieces of what volumes, but it is possible with some
forethought and planning.
I'm really not familiar with FLEX-ES, however.. if it's a form of disk
striping beneath the Linux VM, you're already starting to addres spindle
contention :)
- Ian C. Blenke <ian@blenke.com> <icblenke@nks.net>
http://ian.blenke.com
This archive was generated by hypermail 2.1.3 : Fri Aug 01 2014 - 19:56:33 EDT