Re: RHEL 3.0 Cluster + LVM

[Pasi Pirhonen <upi iki fi>]

Hi,


On Wed, Jun 02, 2004 at 10:00:41AM -0400, Chris Purcell wrote:
> I'm about to set up my first cluster on two RHEL AS 3.0 servers.  I
> normally run LVM on all of my servers, but I wasn't sure if I should do so
> on the shared storage.  I created the two shared partitions (20MB raw
> devices) and was wondering if I should run LVM on the rest of the shared
> disks?
> 


Short version: Sure. It at least should work

Given that it was like (long time ago) and the Redhat for me was then
7.3 and Oracle was 9i/RAC .....

But the consept sould hold still. I did have shared SCSI for my
setup and starting from quorums, everything was over LVM. Those were
just mapped as RAW-disks. I even used shared FireWire-box for a while.

With EVMS2 it worked much better as there the modifications migrated on
the fly to the other node(s). LVM didn't do that so well.

For example if i'd have 1GB LV which is mapper fas RAW for one Oracle
tablespace. Then i want to increase it on the fly. LVM can hand this. I
just had difficulties getting another node(s) to see the change on the
fly. Remember it being down to de-/re-activating the VG which the LV
resided. EVMS2 didn't suffer such problems.

Oracle itself didn't suffer such anomalities. IIRC one could just
expand the tablespace on RAC as it's actually testing if it's able to
expand. Another node(s) just showing 'wrong info' was the side effect.
I may remember this wrong tho.



-- 
Pasi Pirhonen - upi iki fi - http://iki.fi/upi/